Data file: AGRIB57

Title: Analysis over years

Function: FACTOR

Experiment Model Number 15:

One Factor Randomized Complete Block Design Combined over
Locations (or Combined over Years)
Data case no. 1 to 90.

Factorial ANOVA for the factors:

Location (Var 1: Site) with values from 1 to 3
Replication (Var 3: Rep) with values from 1 to 3
Factor A (Var 2: Geno) with values from 1 to 10

Variable 4: PH

Grand Mean = 94.114 Grand Sum = 8470.300 Total Count = 90

T A B L E O F M E A N S

1 3 2 4 Total
-------------------------------------------------------
1 * * 94.353 2830.600
2 * * 91.423 2742.700
3 * * 96.567 2897.000
-------------------------------------------------------
* * 1 97.211 874.900
* * 2 95.522 859.700
* * 3 94.000 846.000
* * 4 90.322 812.900
* * 5 95.411 858.700
* * 6 91.256 821.300
* * 7 96.233 866.100
* * 8 88.867 799.800
* * 9 93.811 844.300
* * 10 98.511 886.600
-------------------------------------------------------
1 * 1 99.467 298.400
1 * 2 97.833 293.500
1 * 3 96.133 288.400
1 * 4 90.600 271.800
1 * 5 95.133 285.400
1 * 6 87.167 261.500
1 * 7 98.600 295.800
1 * 8 85.100 255.300
1 * 9 95.700 287.100
1 * 10 97.800 293.400
2 * 1 91.133 273.400
 2 * 2 93.500 280.500
2 * 3 88.133 264.400
2 * 4 88.733 266.200
2 * 5 94.467 283.400
2 * 6 92.633 277.900
2 * 7 91.933 275.800
2 * 8 86.433 259.300
2 * 9 89.000 267.000
2 * 10 98.267 294.800
3 * 1 101.033 303.100
3 * 2 95.233 285.700
3 * 3 97.733 293.200
3 * 4 91.633 274.900
3 * 5 96.633 289.900
3 * 6 93.967 281.900
3 * 7 98.167 294.500
3 * 8 95.067 285.200
3 * 9 96.733 290.200
3 * 10 99.467 298.400
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 399.376 199.688 0.5522
-3 Error 6 2169.580 361.597
4 Factor A 9 785.442 87.271 1.2680
0.2755
5 LA 18 425.379 23.632 0.3434
-7 Error 54 3716.613 68.826
--------------------------------------------------------------------------
---
Total 89 7496.391
--------------------------------------------------------------------------
---

Coefficient of Variation: 8.81%

s_ for means group 1: 3.4718 Number of Observations: 30

y

s_ for means group 4: 2.7654 Number of Observations: 9

y

s_ for means group 5: 4.7898 Number of Observations: 3

y
==========================================================================
===

Variable 5: PDL

Grand Mean = 19.093 Grand Sum = 1718.400 Total Count = 90

T A B L E O F M E A N S

1 3 2 5 Total
-------------------------------------------------------
1 * * 19.037 571.100
2 * * 18.060 541.800
3 * * 20.183 605.500
-------------------------------------------------------
* * 1 19.356 174.200
* * 2 20.544 184.900
* * 3 19.611 176.500
* * 4 18.867 169.800
* * 5 19.922 179.300
* * 6 18.489 166.400
* * 7 17.722 159.500
* * 8 18.178 163.600
* * 9 18.589 167.300
* * 10 19.656 176.900
-------------------------------------------------------
1 * 1 19.700 59.100
1 * 2 19.333 58.000
1 * 3 21.133 63.400
1 * 4 14.933 44.800
1 * 5 20.600 61.800
1 * 6 18.233 54.700
1 * 7 16.600 49.800
1 * 8 20.900 62.700
1 * 9 20.767 62.300
1 * 10 18.167 54.500
2 * 1 18.833 56.500
2 * 2 20.933 62.800
2 * 3 18.633 55.900
2 * 4 19.100 57.300
2 * 5 15.433 46.300
2 * 6 17.300 51.900
2 * 7 20.167 60.500
2 * 8 16.100 48.300
2 * 9 16.433 49.300
2 * 10 17.667 53.000
3 * 1 19.533 58.600
 3 * 2 21.367 64.100
3 * 3 19.067 57.200
3 * 4 22.567 67.700
3 * 5 23.733 71.200
3 * 6 19.933 59.800
3 * 7 16.400 49.200
3 * 8 17.533 52.600
3 * 9 18.567 55.700
3 * 10 23.133 69.400
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 67.773 33.886 2.3414
0.1772
-3 Error 6 86.837 14.473
4 Factor A 9 61.516 6.835 0.4687
5 LA 18 301.181 16.732 1.1474
0.3362
-7 Error 54 787.449 14.582
--------------------------------------------------------------------------
---
Total 89 1304.756
--------------------------------------------------------------------------
---

Coefficient of Variation: 20.00%

s_ for means group 1: 0.6946 Number of Observations: 30

y

s_ for means group 4: 1.2729 Number of Observations: 9

y

s_ for means group 5: 2.2047 Number of Observations: 3

y

==========================================================================
===
Variable 6: NTPP

Grand Mean = 5.571 Grand Sum = 501.350 Total Count = 90

T A B L E O F M E A N S

1 3 2 6 Total
-------------------------------------------------------
1 * * 5.389 161.680
2 * * 5.525 165.740
3 * * 5.798 173.930
-------------------------------------------------------
* * 1 6.481 58.330
* * 2 5.778 52.000
* * 3 5.683 51.150
* * 4 5.567 50.100
* * 5 5.284 47.560
* * 6 5.281 47.530
* * 7 5.288 47.590
* * 8 5.343 48.090
* * 9 5.307 47.760
* * 10 5.693 51.240
-------------------------------------------------------
1 * 1 6.313 18.940
1 * 2 7.257 21.770
1 * 3 6.170 18.510
1 * 4 6.567 19.700
1 * 5 4.587 13.760
1 * 6 4.377 13.130
1 * 7 4.733 14.200
1 * 8 4.333 13.000
1 * 9 4.300 12.900
1 * 10 5.257 15.770
2 * 1 7.253 21.760
2 * 2 5.733 17.200
2 * 3 5.867 17.600
2 * 4 4.810 14.430
2 * 5 4.710 14.130
2 * 6 5.567 16.700
2 * 7 4.477 13.430
2 * 8 5.520 16.560
2 * 9 4.910 14.730
2 * 10 6.400 19.200
3 * 1 5.877 17.630
3 * 2 4.343 13.030
3 * 3 5.013 15.040
3 * 4 5.323 15.970
3 * 5 6.557 19.670
3 * 6 5.900 17.700
3 * 7 6.653 19.960
3 * 8 6.177 18.530
3 * 9 6.710 20.130
3 * 10 5.423 16.270
 -------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 2.596 1.298 1.8887
0.2311
-3 Error 6 4.123 0.687
4 Factor A 9 11.401 1.267 0.5611
5 LA 18 56.758 3.153 1.3968
0.1714
-7 Error 54 121.905 2.258
--------------------------------------------------------------------------
---
Total 89 196.782
--------------------------------------------------------------------------
---

Coefficient of Variation: 26.97%

s_ for means group 1: 0.1514 Number of Observations: 30

y

s_ for means group 4: 0.5008 Number of Observations: 9

y

s_ for means group 5: 0.8675 Number of Observations: 3

y

==========================================================================
===

Variable 7: LAI

Grand Mean = 24.523 Grand Sum = 2207.060 Total Count = 90

T A B L E O F M E A N S

1 3 2 7 Total
-------------------------------------------------------
 1 * * 24.552 736.560
2 * * 24.638 739.130
3 * * 24.379 731.370
-------------------------------------------------------
* * 1 25.437 228.930
* * 2 24.773 222.960
* * 3 25.107 225.960
* * 4 24.348 219.130
* * 5 24.377 219.390
* * 6 23.107 207.960
* * 7 23.677 213.090
* * 8 25.089 225.800
* * 9 25.016 225.140
* * 10 24.300 218.700
-------------------------------------------------------
1 * 1 25.400 76.200
1 * 2 25.117 75.350
1 * 3 25.500 76.500
1 * 4 24.610 73.830
1 * 5 25.453 76.360
1 * 6 22.303 66.910
1 * 7 23.350 70.050
1 * 8 24.957 74.870
1 * 9 24.610 73.830
1 * 10 24.220 72.660
2 * 1 24.577 73.730
2 * 2 25.200 75.600
2 * 3 24.790 74.370
2 * 4 24.663 73.990
2 * 5 23.023 69.070
2 * 6 22.573 67.720
2 * 7 25.123 75.370
2 * 8 25.510 76.530
2 * 9 25.627 76.880
2 * 10 25.290 75.870
3 * 1 26.333 79.000
3 * 2 24.003 72.010
3 * 3 25.030 75.090
3 * 4 23.770 71.310
3 * 5 24.653 73.960
3 * 6 24.443 73.330
3 * 7 22.557 67.670
3 * 8 24.800 74.400
3 * 9 24.810 74.430
3 * 10 23.390 70.170
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 1.042 0.521 0.8309
-3 Error 6 3.761 0.627
4 Factor A 9 41.626 4.625 2.3148
0.0278
5 LA 18 44.293 2.461 1.2316
0.2708
-7 Error 54 107.895 1.998
--------------------------------------------------------------------------
---
Total 89 198.616
--------------------------------------------------------------------------
---

Coefficient of Variation: 5.76%

s_ for means group 1: 0.1446 Number of Observations: 30

y

s_ for means group 4: 0.4712 Number of Observations: 9

y

s_ for means group 5: 0.8161 Number of Observations: 3

y
 Assessment of Academic
Program/Department Program
Teams (PT) at Department level,
Ghazi University Dera Ghazi Khan
Faculty of Sciences

Department of Statistics

Sr # Name Designation
1 Mr. Niyyar Munir Lecturer Convener
2 Mr. Mujahid Hussain Assistant Professor Member
3 Mr. Muhammad Ali Lecturer (Adhoc) Member

construct grouped frequency distribution for the following data

Classes/Groups Tally f
36 39 38
36-38 //
41 42 43
39-41 ////
43 45 47
42-44 //// //
47 49 47
45-47 //// //// //
44 46 45
48-50 ////
39 42 39
51-53 //
44 47 45
Total 45
46 49 48

51 52 49

46 47 47

40 43 39

43 45 43

48 49 47
 45 51 47

46 48 45

Design and Analysis of Experiment

What is science?

The study of nature and its behavior.

What is nature?

The growing of seed

Rain

Mountains

Plants

Human

What are scientific methods?

Application of logic and objectivity to the understanding of a phenomena,

Logic and Philosophy

Y = a +bx

What is experiment?

A planned inquiry

Random experiment

When we sow a seed

Whether it grows or not

This is unpredictable.

There are different results.

You sow three seed of a particular flower.

What is the result.

Which show different results.

What is the purpose of experiment?

Essentials of scientific methods are examination of what is already known and formulation of hypothesis
which may be put to experimental test, examination, and objective verification. In short, we can say that
a theory is such only based on the facts, or a fact is such only based on theory

What is theory

Theory is the verified result basis on some facts or assumptions

Repetition Machines Difference

CASIO Sharp
1 30 14 16
2 21 21 0
3 22 5 17
4 22 13 9
5 18 13 5
6 29 17 12
7 16 7 9
8 12 16 -4
9 23 8 15
10 23 24 -1
Average 21.6 13.8 7.8

This table shows the speed of two machines for the time required in seconds for computing the sum of
square

Is there any difference between the average of two machines

2 2
s1 =29.6 , s2 =36.6

x 1=21.6 , x 2=13.8

n1 = 10, n2 = 10

H0: µ1=µ2 H0: µ1−µ2=0

H1: µ1 ≠ µ2 , H1: µ1−µ2 ≠ 0

α= 0.05

Test Statistics:

( x 1−x 2) −(μ 1−μ2) ( 21.6−13.8 )−(0)

t= t= 7.8

√ √1 1 , t= ( 5.75 ) (0.4472) , t=3.03

Test Statistics: 1 1 ,
sp + (5.75) +
n1 n 2 10 10
Where, s p is called as pooled standard deviation

√
s p= (n¿¿ 1−1)s 1 +
(n¿¿ 2−1)s 22
2
n1+ n2−2
2
¿ ¿, s p=
√
10(229.6)+10(36.6) s =5.75
18
, p ,

2 (Σ x 1)
2
2 1 1 (216) ¿ 1 (4932−4665.6)
s1 = (Σ x1 − ),¿ (4932− ),
n1−1 n1 10−1 10 9
2
s1 =29.6
2
2 (Σ x 2)
2
2 1 1 (138) ¿ 1 (2234−1904.4)
s2 = (Σ x2 − ) ,¿ (2234− ),
n2−1 n2 10−1 10 9
2
s2 =36.6
Critical Region

If H0: µ1=µ2 H1: µ1 ≠ µ2, Two-sided test

i.e. t 0.05 ,18<−t , and t 0.05 ,18> ¿ t,

2 2

t 0.025 ,18<−2.4450, and t 0.025 ,18> ¿ 2.4450

If H0: µ1=µ2 H1: µ1 ≠ µ2, Two-sided test

i.e. t 0.05 ,23<−t , and t 0.05 ,23> ¿ t,

2 2

t 0.025 ,23<−2.069, and t 0.025 ,23> ¿ 2.069

Accept the null hypothesis

Replications Machines Difference

CASIO Sharp
1 14 30 -16
2 21 21 0
3 5 22 -17
4 13 22 -9
5 13 18 -5
6 17 29 -12
7 7 16 -9
8 16 12 4
9 8 23 -15
10 24 23 1
Average 13.8 21.6 -7.8
Experimental Material

Experimental unit

What is Treatment

The smallest division of experimental material is called as experimental unit

A unit may be a human, animal, plant, test tube, dish, pot field of som specific size etc.

Replication

Repetition of the treatment to the experimental units under same conditions

How many experimental units should be use (No of treatments X No. of replications)

r x t= no of experimental units = n

n=rxt

The role of Statistics in experimentation

Reasoning is based on two types

Deductive reasoning:- in this consequences in the premises deduced and verified by observations and
experimentatioin.

Indective reasoning:- it is a process of drawing general conclusions from a particular set of instances,
depends upon a fair sample of the the population about which make statements.

Basically theory of statistics based on three functions

1. Description
2. Analysis
3. Prediction

Descriptive statistics reduce the mass of the data and concise the information, it is a deductive
procedure, if we generalize it for the whole population then it is inductive.

Analysis is more frequently used method to tell about a particular distribution or two compare two
samples whether they come from same population

Prediction is for unknown values of the parameter

The designs of experiment:-

The general purpose of scientific research is to formulate the hypothesis and to verify them. Collection
of data in design of experiment is essential. The consideration of the design of experiment to provide
evidence for a hypothesis, whether the hypothesis is capable of being verified or not. As we discussed
earlier in the example of machine A and machine B.

The word experiment means “a trial made to confirm or disprove something doubtful under
some conditions”. There are two types of experiments, absolute experiments and comparative
experiments, sample survey is an absolute experiment and on the other hand comparative experiment is
an experiment in which two or more treatments are compared in their effects.

The principles of experimental designs: -

A statistically designed investigation consists of the following steps.

1. Statement of the problem.

2. Formulation of hypothesis.
3. Planning or designing an experiment.
4. Examination of possible outcomes and reference back to the reasons (Review of literature) for
the inquiry to be sure or not.
5. Consider possible results from statistical point of view.
6. Performance of an experiment.
7. Application of statistical techniques.
8. Drawing conclusions.
9. Evaluation with other investigations on the same problem.

When we conduct an experiment there are three basic principles

1. Randomization
2. Replication
3. Local control

Before starting an experiment after formation of hypothesis the next step is to design an experiment
which has following different phases

1. Select appropriate and homogeneous material to test (e.g. Laboratory equipments, Green
housebmaterial, Field experiment etc)
2. Decide no of factors and their treatments along with their characters/parameters/variables (e.g.
Fertilizer is a factor and different doses of fertilizer such as 50 kg, 100 kg, 150 kg per acer, are
the treatments of a factor) in other words treatments of a factor are also called levels. If an
experiment based on a sigle factor it is called as siangle factor experiment.
3. Decide the no parameters for aaalysis (e.g. if we want to check the effect of fertilizer on wheat
or on rice then the grain yield, no of grains per plant plant height, no of straws etc are the
parameters).
4. Select the procedure/scale to measure those characters
5. Specify the statistical methods to determine whether the measurements support the hypothesis
or not.

Now to design the experiment we have to follow basic principles of experimental designs

1. Randomization:- Let we have three treatments namely T1, T2, T3 and we have to determine
which experimental unit receive T1, which experimental unit receive T2, and which experimental
unit receive T3 without any personal interest. It means the allocation of the treatments to the
experimental units independently to avoid bias. For this purpose (avoid from bias)
randomization is very important way to achieve this goal. For three treatments we need three
experimental units and we choose one unit or one treatment randomly. Through the
 randomization procedure the reliability of the experiment increase. There are number of ways
that randomization can be achieved. Romdom no table, calculator, tags, gold fish bowl method,
statistical software.
2. Replication:- To find the effect of treatment between to experimental units which are treated
alike we repeat the treatments more than one. The repetition of the treatments is used to
measure the experimental error. There are following uses of replications
i. To secure more accurate estimates of experimental error
ii. To increase the precision and decrease the experimental error
2
σ
iii. To obtain more precise estimates of the mean effect of a treatment σ x 2=
n
3. Local control:- It is very important for a researcher that he use all the possible methods to
minimize the experimental error. As we know that the experimental error produced due to
extraneous factors other than treatments. For this purpose we use local control, a term used for
blocking, grouping, and balancing the homogenous experimental units. The most commonly
used technique of local control is blocking.
Blocking:- There is two types of blocking, one-way blocking and two-ways blocking based on the
circumstances and atmosphere and situations where the experiment is going to be conducted.
Blocking means that like experimental units or homogenous experimental units should be
collected together. Which form a homogenous group. The main purpose of the local control is
to increase the efficiency of the experiment by decreasing the experimental error. In blocking
we put homogenous units together and the size of the block depends upon the treatments. A
block should contain all the treatments. In other word we can say that if there are three
treatment then a block should consists of three experimental units. And assign the treatments
to the units randomly. The important point is that it is not necessary that the units in a block are
adjacent but it is necessary that they should be homogenous.
 There are basically two types of designs
Randomized designs
Systematic designs
Randomized designs are those type of designs in which all treatments are applied to the
experimental units completely at random without any personal interest.
There are different types of experiments
i. Single factor experiments
ii. Two factor experiments
iii. Multi factor experiments

In single factor experiments only one factor is involved or more than one factors with only one
level of treatment

For example we want to compare different irrigation levels (Flood irrigation, sprinkle, and drip
irrigation)

Or we want to compare different doses of fertilizers on a specific crop (what is the effect of 50
kg urea on one acre 100 kg urea on one acre 150 kg urea on one acre

Apply sodium, calcium and nitrogen

In single factor experiments there are two situations

a. Complete block designs

b. Incomplete block designs

1. Completely randomized designs

Completely randomized designs (CRD) is defined as iin which the treatments are assigned to the
experimental units completely at random. Randomization is done without any restriction. It is a most
flexible design as any number treatments and any number replications can be used. The total number of
experimental units in this desn is multiple of no of treatments and no of replications (n = rxt). If we have
three treatments T1, T2, T3 and four replications R1, R2, R3, R4 then total no experimental units will be
12. (n = 4 x 3 = 12)

Application of CR-Designs

These designs are applied in those situations where

The experimental units are homogeneous

Experiments are small, and

Some of the experimental units are destroyed and failed to respond

The most appropriate situations for CR-Designs are laboratory experiments, pot experiments and green
house experiments
Layout plan for CR-Designs

Let we have four treatments with three replications for each treatment

The treatments are T1, T2, T3, and T4

The replications are R1, R2, and R3

It means we repeat each treatment three times so we need twelve experimental units

AT4 BT1 CT3 DT4

HT1 GT2 FT1 ET3
IT2 JT2 KT3 LT4

Treatments Random nos

T4 001
T1 010
T3 065
T4 112
T1 640
T2 654
T1 666
T3 772
T2 888
T2 912
T3 999
T4

Comparison of treatments and analysis

One-way ANOVA is used to compare the treatment means

The sketch for ANOVA is

H0 : All the treatment are equal

H1 : Atleast two means are not equal

SOV df SS MS Fcal F tab/P-value

Treatments 4-1=3 282.25 94.08 11.29** 4.05/0.003
Error 12-4=8 66.67 8.33
Total 12-1=11 348.92
Find correction factor

CF = (GT)2/n = 92050.1
 t r
Total SS=∑ ❑ ∑ y ij 2−CF = 348.917
j=1 i=1

∑ T j2 = 282.25
Treatment SS= j=1 −CF
r
Unbiased variance
t n
2
∑ xj
1 j =1
2
( ∑ x i)
s2= [ − i=1
]
n−1 ❑ n
Error SS = Total SS- Treatment SS = 66.67

There was a significant difference between the treatment means

I f the null hypothesis is rejected then use pair comparison test

Least significant difference test (LSD-Test)

Duncan’s multiple range test (DMR-Test)

Tuckey’s test

Schefe’s test

Contrast

What is p-value, How it can be calculated for z-test, t-test, chisquare-test, and F-test

CR-Designs with unequal replications

If there are unequal replications in CR-Designs then only treatment SS will be calculated as follow
2 2 2
T 1 T2 T 3
Treatment SS=[ + + ]−CF = 282.25
r1 r2 r 3

All other SS will be calculated in the same manners as in the equal replication case

Let we have four treatments with three replications for T1, two replications for T2 and three replications
for T3, and three replications for T4.

The treatments are T1, T2, T3, and T4

The replications are R1, R2, and R3

H0 : All the treatment are equal

H1 : Atleast two means are not equal

SOV df SS MS Fcal F tab/P-value
Treatments 4-1=3 269.742 89.91 9.66 4.35/0.005
Error 11-4=7 65.17 9.31
Total 11-1=110 334.909
Randomized Complete Block Designs (RCBD)

It is the design in which hpmpgeneous experimental units are combined I a group called as Block, in such
a wat that a block contains complete set of treatments, these designs are not flexible as that of CR-
Dsigns. The restriction in these design is that a single treatment occurs only once in a single block. These
designs are most frequently used designs.

Application

Mostly these designs are applied in field experiments, A field is ditrbuted in block x treatment
experimental units (N = B x T)

H0 : All the treatment are equal

H1 : Atleast two means are not equal

H0 : All the Block means are equal

H1 : Atleast two block means are not equal

SOV df SS MS Fcal F tab/P-value

Block B-1 = 2 20.67 10.33 1.35 0.3285
Treatments t-1 = 3 94.08 31.36 4.09 0.0617
Error (B-1)(t-1) = 6 46.00 7.67
Total N-1 = 11 348.92
Find correction factor

CF = (GT)2/n
t r
Total SS=∑ ❑ ∑ y ij 2−CF
j=1 i=1

∑ T j2
Treatment SS= j=1 −CF
r
b

∑ Bi2
Block SS= r =1 −CF
t
Error SS = Total SS- Treatment SS – Block SS

Accept the null hypothesis for treatments and blocks

 Statistical model for CR-Designs
The basic Statistical model for CR-Designs is

Y ij =μ+ α j +∈ij
k
E(Total SS )=r ∑ α j2 + ( rk −1 ) σ 2
j=1

r k
Total SS=∑ ∑ Y ij 2−CF
i=1 j=1

r k r k

∑ ∑ Y ij2=∑ ∑ (μ+ α j+ ∈ij )2

i=1 j=1 i=1 j =1

r k r k

∑ ∑ Y ij2=∑ ∑ (μ2 + α j2 +∈ij2+ ¿¿2 μ α j +2 μ ∈ij+2 α j ∈ij )¿ ¿

i=1 j=1 i=1 j =1

r k k r k k r k r k
2
∑∑ Y ij
2
=r . k μ +r ∑ ❑α j + ∑ ∑ ∈ij +2 μr ∑ α j +2 μ ∑ ∑ ∈ij +2 ∑ ∑ α j ∈ij
2 2

i=1 j=1 j=1 i=1 j=1 j =1 i=1 j=1 i=1 j=1

By applying expectation on both sides

r k k r k r k
2
E(∑ ∑ Y ¿=E (r . k μ + r ∑ ❑ α j + ∑ ∑ ∈ij +2 ∑ ∑ α j ∈ij )
2 2 2
ij
i=1 j=1 j=1 i=1 j=1 i=1 j =1

r k k
E(∑ ∑ Y ij ¿=r . k μ +r ∑ α j +r . k . σ
2 2 2

i=1 j=1 j=1

So
k
E(Total SS )=r . k μ 2+ r ∑ α j2❑+ r . k . σ 2 −(r . k . μ 2+ σ 2 )
j=1

k
E(Total SS )=r . k μ 2+ r ∑ α j2❑+ r . k . σ 2 −r . k . μ2−σ 2 ¿
j=1

k
E(Total SS )=r ∑ α j2❑+r . k . σ 2−σ 2 ¿
j=1

k
E(Total SS )=r ∑ α j2❑+(r . k .−1)σ 2
j=1

N2 N2 N3 N4 N5
 R1 88.4 92.4 96.8 101.2 97.3
R2 83.4 96.7 97.1 99.8 96.5
R3 85.3 93.6 98.3 103.4 94.7
Total 257.1 282.7 292.2 304.4 288.5

Now we find Sum of Square for treatments

The basic Statistical model for CR-Designs is

Y ij =μ+ α j +∈ij
k
E(Treatment SS)=r ∑ α j2+ ( k−1 ) σ 2
j=1

∑ Y . j2
Treatment SS= j=1 −CF
r
r r
Y . j=∑ Y ij=∑ (μ+ α j+∈ij )
i=1 i=1

r
Y . j=rμ +rα j + ∑ ∈ij
i=1

Squaring both side

2
Y . j =¿
2 2 2 2 2
Y . j =r μ +r α j+ ¿ ¿
Apply summation on both sides
k k

∑ Y . j 2= ∑ ¿ ¿
j=1 j=1

k k

∑ Y . j =r 2 2 2
k μ +r 2
∑ α 2 j +¿
j=1 j=1

Divide by r on both sides

k

∑ Y . j2 1
j =1
= ¿
r r
 k

∑ Y . j2 k
1
j =1
=rk μ + r ∑ α 2 j+ ¿
2
r j=1 r
As we know that

¿
So
k

∑ Y . j2 1
k
2
❑ ❑ r k k r k k r k
j =1
=rk μ + r ∑ α j+ ( ∑ ∑ ∈ij + ∑ ∑ ∈ij ∈gh)+2 rμ ∑ α j +2 μ ∑ ∑ ∈ij +2 ∑ α j ∑ ∑ ∈ij
2 2
r j=1 r i=1 j=1 i ≠ g j≠ h j=1 i=1 j=1 j=1 i=1 j=1

Now taking expectation on both sides

k

∑ Y . j2 1
k
2
❑ ❑ r k k r k k r k
E [ j=1 ]=E[rk μ2 +r ∑ α 2 j + ( ∑ ∑ ∈ij +∑ ∑ ∈ij ∈gh )+2 rμ ∑ α j +2 μ ∑ ∑ ∈ij +2 ∑ α j ∑ ∑ ∈ij ]
r j=1 r i=1 j=1 i ≠ g j ≠h j=1 i=1 j=1 j=1 i =1 j=1

∑ Y . j2 1
k
E[ j=1
]=rk μ +r ∑ α 2 j + (rk σ 2+ 0)+ 2rμ (0)+2 μ (0)+2(0)
2
r j =1 r
k

∑ Y . j2 k
E[ j=1
]=rk μ +r ∑ α 2 j +(k σ 2)
2
r j =1

E(Treatment SS)=E ¿
2 2
E(CF )=rk μ +σ
k
E ( Treatment SS )=rk μ 2+ r ∑ α 2 j+ ( k σ 2 ) −(rk μ 2+ σ 2)
j=1

k
E(Treatment SS)=r ∑ α j2+ ( k−1 ) σ 2
j=1

Now for error sum of squares

E(Error SS) = E(Total SS) – E(Treatment SS)

k
E(Total SS )=r ∑ α j2❑+(r . k .−1)σ 2
j=1

k
E(Treatment SS)=r ∑ α j2+ ( k−1 ) σ 2
j=1
 k k
E ( Error SS )=r ∑ α 2
j ❑ + r . k . σ −σ −(r ∑ α j2 +k σ 2−σ 2)
2 2

j =1 j=1

k k
E ( Error SS )=r ∑ α j2❑+ r . k . σ 2−σ 2−r ∑ α j2−k σ 2+ σ 2
j =1 j=1

2 2
E ( Error SS )=r . k . σ −k σ ¿
2
E ( Error SS )=(r . k .−k )σ
2
(r . k .−k )σ
E ( Error MS )=
rk −k
2
E ( Error MS )=σ
What is the difference between fixed effect models random effect models and mixed effect models

Explain with examples and write down their applications in different fields of study
 Random Effect Model
As we know that treatments are the samples of a populationfrom which we make inferences about a
population. If a researcher randomly select a set of treatments from a large number of treatments then
we can say that the factor is random for example there are a number of pain killers then researcher
wants to test the effectiveness of of any three of them which are selected randomly then the model
applied is random effect model

Statistical Model

Y ij =μ+ α j +∈ij

I = 1, 2 , …, r (Replications))

J = 1, 2, …, k (Treatments)

With the assumptions that

2
∈ij NII (0 , σ )

E ( ∈2ij )=σ 2, E ( ∈ij )=0, E ( ∈ij ∈gh ) =0 , wherei ≠ g ,∧ j ≠ h

As it is a random effect model, so assumptions for treatment effects are

E ( α 2 j ) =σ 2α , E ( α j ) =0, E ( α j α m )=0 , where j≠ m

2
And α j NII (0 , σ α )

Also α j and ∈ij are independent

Prove that

E(MSE) = σ 2

Correction Factor

C.F = G2/rk
r k
G = ∑ ∑ Y ij
i=1 j=1

r k
G = ∑ ∑ (μ+ α j +∈ij )
i=1 j=1

k r k
G = rk μ+r ∑ α j+ ∑ ∑ ∈ij
j=1 i=1 j=1
Squaring both sides
k r k
2
G2 ¿(rk μ+r ∑ α j+ ∑ ∑ ∈ij )
j=1 i=1 j=1

( ) ( )
k r k k r k k r k
2 2
G2=(rk μ)2+(r ∑ α j ) +( ∑ ∑ ∈ij ) +2(rk μ ¿ r ∑ α j +2 ( rk μ ) ∑ ∑ ∈ij +2(r ∑ α j )(∑ ∑ ∈ij)
j=1 i=1 j=1 j=1 i=1 j =1 j=1 i=1 j=1

Now use the assumptions

k k k
2
( ∑ α j) =∑ ∝2j + ∑ α j α m
j=1 j=1 j≠m

r k r k r k
2
( ∑ ∑ ∈ij ) =∑ ∑ ∈ + ∑ ∑ ∈ij ∈ gh 2
ij
i=1 j=1 i=1 j =1 i≠g j≠h

By divide rk and taking expectation on both sides we get

k k r k r k
G2 r 1
E[ ]=rk μ2 + E( ∑ ∝2j + ∑ α j α m )+ E( ∑ ∑ ∈2ij + ∑ ∑ ∈ij ∈gh )+2(
rk k j=1 j ≠m rk i=1 j=1 i≠g j≠h

( ) ( )
k r k k r k
r μ¿ E ∑ α j +2 ( μ ) E ∑ ∑ ∈ij + 2k E(∑ α j )(∑ ∑ ∈ij )
j=1 i=1 j=1 j=1 i=1 j=1

[ ]
2
G 2 2 2
E(CF )=E =rk μ + r σ α +σ
rk
Total Sum of Square

E(Total SS) = ?
r k
Total SS = ∑ ∑ Y ij −CF
2

i=1 j=1

r k

∑ ∑ Y 2ij=?
i=1 j=1

We know that

Y ij =μ+ α j +∈ij
2 2
Y ij ¿ ( μ+α j +∈ij )
2 2 2 2
Y ij =μ +(α ¿¿ j) +(∈ij ) +2 μ α j +2 μ ∈ij +2 α j ∈ij ¿
r k r k

∑ ∑ Y =∑ ∑ ¿ ¿ ¿ 2
ij
i=1 j=1 i=1 j=1
 r k k

∑∑ Y 2
ij =rk μ +r ∑ (α ¿¿ j)2 (¿❑)❑ ¿ ¿
2

i=1 j=1 j=1

 RANDOMIZED COMPLETE BLOCK
DESIGNS WITH FIXED EFFECT
MODELS
Statistical Model

The basic Statistical model for RCB-Designs is

Y ij =μ+ α j + β i +∈ij

I = 1, 2 , …, r (Blocks)

J = 1, 2, …, k (Treatments)

With the assumptions that

2
∈ij NII (0 , σ )

E ( ∈2ij )=σ 2, E ( ∈ij )=0, E ( ∈ij ∈gh ) =0 , wherei ≠ g ,∧ j ≠ h

In case of fixed effect models the treatment effect and block effect are fixed such that

E ( α j ) =0, E ( β i )=0
k r

∑ α j=0, ∑ β i=0
i=1 j=1

Correction factor
2
G
CF=
rk
r k r k k r k
G=∑ ∑ Y ij = ∑ ∑ (μ+ α j + β i +∈ij ) = rk μ+r ∑ α j+ k ∑ r β i +∑ ∑ ∈ij
i=1 j=1 i=1 j=1 j=1 i =¿ ¿ i=1 j=1

r k
, G=rk μ+ ∑ ∑ ∈ij
i=1 j=1

Squaring both sides and divide by rk

r k
2
G =(rk μ+ ∑ ∑ ∈ij )
2

i=1 j=1

r k r k
2
G2=r 2 k 2 μ2 +( ∑ ∑ ∈ij ) +2 rk μ ∑ ∑ ∈ij
i =1 j=1 i=1 j=1

As we know that
 r k r k r k
2
( ∑ ∑ ∈ij ) =∑ ∑ ∈ + ∑ ∑ ∈ij ∈ gh 2
ij
i=1 j=1 i=1 j =1 i≠g j≠h

So,
r k r k
G =r k μ + ∑ ∑ ∈ + ∑ ∑ ∈ij ∈gh
2 2 2 2 2
ij
i=1 j=1 i≠ g j ≠h

Divide by rk on both sides and apply expectation

[ ]
2 r k r k
G 1
E [ CF ] =E = E [r k μ + ∑ ∑ ∈ij +∑ ∑ ∈ij ∈gh ]
2 2 2 2
rk rk i=1 j=1 i ≠ g j ≠h

[ ]
2
G 1 2 2 2
E [ CF ] =E = [ r k μ +rk σ ] =rk μ +σ
2 2 2
rk rk
Hence,

E [ CF ] =rk μ + σ
2 2

Total Sum of Square

E(Total SS)
r k
Total SS = ∑ ∑ Y ij −CF
2

i=1 j=1

r k

∑ ∑ Y 2ij=?
i=1 j=1

We know that

Y ij =μ+ α j + β i +∈ij
2 2
Y ij ¿ ( μ+α j + β i +∈ij )
2 2 2
Y ij =μ +(α ¿¿ j) +¿ ¿
r k r k

∑ ∑ Y =∑ ∑ ¿ ¿ ¿ 2
ij
i=1 j=1 i=1 j=1

[∑ ∑ ]
r k k r
E Y =rk μ +r ∑ (α ¿¿ j) +k ∑ (β¿¿ i)2 ¿ ¿
2
ij
2 2

i=1 j=1 j=1 i=1

r k
E[Total SS] =E[ ∑ ∑ Y ij ¿−E[CF]
2

i=1 j=1

k r
E[Total SS] = r ∑ ( α ¿¿ j) +k ∑ ( β ¿¿ i) +(rk −1)σ ¿ ¿
2 2 2

j=1 i=1
Treatment Sum of Squares

E[Treatment SS]
k

∑ Y . j2
Treatment SS= j=1 −CF
r
r r
Y . j=∑ Y ij=∑ (μ+ α j+ βi +∈ij )
i=1 i=1

r r
Y . j=rμ +rα j + ∑ β i + ∑ ∈ij
i=1 i=1

As for fixed effect models

r

∑ β i=0
j=1

So,
r
Y . j=rμ +rα j + ∑ ∈ij
i=1

Squaring both side

2
Y . j =¿
2 2 2 2 2
Y . j =r μ +r α j+ ¿ ¿
Apply summation on both sides
k k

∑ Y . j 2= ∑ ¿ ¿
j=1 j=1

k k

∑ Y . j =r 2 2 2
k μ +r 2
∑ α 2 j +¿
j=1 j=1

Divide by r on both sides

k

∑ Y . j2 1
j =1
= ¿
r r
 k

∑ Y . j2 1
k
j =1
=rk μ + r ∑ α 2 j+ ¿
2
r j=1 r
As we know that

¿,
So
k

∑ Y . j2 1
k
2
❑ ❑ r k k r k k r k
j =1
=rk μ + r ∑ α j+ ( ∑ ∑ ∈ij + ∑ ∑ ∈ij ∈gh)+2 rμ ∑ α j +2 μ ∑ ∑ ∈ij +2 ∑ α j ∑ ∑ ∈ij
2 2
r j=1 r i=1 j=1 i ≠ g j≠ h j=1 i=1 j=1 j=1 i=1 j=1

Now taking expectation on both sides

k

∑ Y . j2 1
k
2
❑ ❑ r k k r k k r k
E [ j=1 ]=E[rk μ2 +r ∑ α 2 j + ( ∑ ∑ ∈ij +∑ ∑ ∈ij ∈gh )+2 rμ ∑ α j +2 μ ∑ ∑ ∈ij +2 ∑ α j ∑ ∑ ∈ij ]
r j=1 r i=1 j=1 i ≠ g j ≠h j=1 i=1 j=1 j=1 i =1 j=1

∑ Y . j2 1
k
E[ j=1
]=rk μ +r ∑ α 2 j + (rk σ 2+ 0)+ 2rμ (0)+2 μ (0)+2(0)
2
r j =1 r
k

∑ Y . j2 k
E[ j=1
]=rk μ +r ∑ α 2 j +(k σ 2)
2
r j =1

E(Treatment SS)=E ¿
2 2
E(CF )=rk μ +σ
k
E ( Treatment SS )=rk μ 2+ r ∑ α 2 j+ ( k σ 2 ) −(rk μ 2+ σ 2)
j=1

k
E(Treatment SS)=r ∑ α j2+ ( k−1 ) σ 2
j=1

Block Sum of Squares

E[Block SS]
r

∑ Y i .2
Block SS= i=1 −CF
k
k k
Y i . =∑ Y ij =∑ (μ +α j + β i+ ∈ij )
j=1 j=1
 k k
Y i . =kμ+ ∑ α j +k β i+ ∑ ∈ij
j=1 j=1

As for fixed effect models

r

∑ α j=0
j=1

So,
k
Y i . =kμ+ k βi + ∑ ∈ij
j=1

Squaring both side

2
Y i . =¿
2 2 2 2 2
Y i . =k μ + k β i +¿ ¿
Apply summation on both sides
r r

∑ Y i. =∑ ¿ ¿ 2

i=1 i=1

r r r k r k r r k r
2
∑ Y i.2=r k 2 μ2 +k 2 ∑ β 2i +(∑ ∑ ∈ij) +2 k 2 μ ∑ βi +2 kμ ∑ ∑ ∈ij +2 k ∑ βi ∑ ∑ ∈ij
i=1 i=1 i=1 j=1 i =1 j=1 i=1 i=1 j=1 i=1

Divide by k, apply properties, an taking expectation on both sides

k

∑ Y . j2 1 2
r r k r k r r k r
j =1
= [r k 2 μ 2+ k 2 ∑ βi2+( ∑ ∑ ∈ij ) +2 k 2 μ ∑ β i +2 kμ ∑ ∑ ∈ij + 2 k ∑ βi ∑ ∑ ∈ij ]
k k i=1 i=1 j=1 i=1 j=1 i=1 i=1 j=1 i=1

As we know that

¿,
So
r

∑ Y i .2 r
E[ i=1
]=rk μ + k ∑ β2i +(r σ 2 )
2
r i =1

E(Block SS)=E ¿
 2 2
E(CF )=rk μ +σ
r
E ( Block SS )=rk μ + k ∑ β2i +(r σ 2 )−(rk μ 2+ σ 2 )
2

i=1

r
E(Block SS)=k ∑ β2i + ( r −1 ) σ 2
i=1
Error Sum of square

E(Error SS) = ?

Error SS = Total SS -Treatment SS – Block SS

E(Error SS) = E(Total SS) – E(Treatment SS) – E(Block SS)

k r
E[Total SS] = r ∑ ( α ¿¿ j) +k ∑ ( β ¿¿ i) +(rk −1)σ ¿ ¿
2 2 2

j=1 i=1

k
E(Treatment SS)=r ∑ α j2+ ( k−1 ) σ 2
j=1

r
E(Block SS)=k ∑ β2i + ( r −1 ) σ 2
i=1

( )
k r k r
E(Error SS) = r ∑ ( α ¿¿ j) +k ∑ ( β ¿¿ i) + ( rk −1 ) σ − r ∑ α j + ( k−1 ) σ −(k ∑ β i + ( r−1 ) σ )¿ ¿
2 2 2 2 2 2 2

j=1 i=1 j=1 i=1

E(Error SS) = ( rk −1 ) σ 2−( ( k−1 ) σ 2) −( ( r−1 ) σ 2 )

E(Error SS) = (r −1)(k−1)σ 2

E(MSE) = σ 2

N2 N2 N3 N4 N5 Total
B1 88.4 92.4 96.8 101.2 97.3 476.1
B2 83.4 96.7 97.1 99.8 96.5 473.5
B3 85.3 93.6 98.3 103.4 94.7 475.3
Total 257.1 282.7 292.2 304.4 288.5 1424.9

X and Y

Var (X)

Var (Y)

Cov (X, Y)

X 1 and X 2

Var ( X 1 ) = S11

Var ( X 2 )= S22
Cov ( X 1 , X 2 ¿

Cov (1, 2 ¿= S12

2
(a+ b+c +d )
=a2+b2+c2+d2+2ab+2ac+2ad+2bc+2bd+2cd

)
 RANDOMIZED COMPLETE BLOCK
DESIGNS WITH RANDOM EFFECT
MODELS
Statistical Model

The basic Statistical model for RCB-Designs is

Y ij =μ+ α j + β i +∈ij

I = 1, 2 , …, r (Blocks)

J = 1, 2, …, k (Treatments)

With the assumptions that

2
∈ij NII (0 , σ )

E ( ∈2ij )=σ 2, E ( ∈ij )=0, E ( ∈ij ∈gh ) =0 , wherei ≠ g ,∧ j ≠ h

In case of fixed effect models the treatment effect and block effect are fixed such that

E ( α j ) =0, E ( β i )=0
k r

∑ α j=0, ∑ β i=0
i=1 j=1

Correction factor
2
G
CF=
rk
r k r k k r k
G=∑ ∑ Y ij = ∑ ∑ (μ+ α j + β i +∈ij ) = rk μ+r ∑ α j+ k ∑ r β i +∑ ∑ ∈ij
i=1 j=1 i=1 j=1 j=1 i =¿ ¿ i=1 j=1

r k
, G=rk μ+ ∑ ∑ ∈ij
i=1 j=1

Squaring both sides and divide by rk

r k
2
G =(rk μ+ ∑ ∑ ∈ij )
2

i=1 j=1

r k r k
2
G2=r 2 k 2 μ2 +( ∑ ∑ ∈ij ) +2 rk μ ∑ ∑ ∈ij
i =1 j=1 i=1 j=1

As we know that
 r k r k r k
2
( ∑ ∑ ∈ij ) =∑ ∑ ∈ + ∑ ∑ ∈ij ∈ gh 2
ij
i=1 j=1 i=1 j =1 i≠g j≠h

So,
r k r k
G =r k μ + ∑ ∑ ∈ + ∑ ∑ ∈ij ∈gh
2 2 2 2 2
ij
i=1 j=1 i≠ g j ≠h

Divide by rk on both sides and apply expectation

[ ]
2 r k r k
G 1
E [ CF ] =E = E [r k μ + ∑ ∑ ∈ij +∑ ∑ ∈ij ∈gh ]
2 2 2 2
rk rk i=1 j=1 i ≠ g j ≠h

[ ]
2
G 1 2 2 2
E [ CF ] =E = [ r k μ +rk σ ] =rk μ +σ
2 2 2
rk rk
Hence,

E [ CF ] =rk μ + σ
2 2

Total Sum of Square

E(Total SS)
r k
Total SS = ∑ ∑ Y ij −CF
2

i=1 j=1

r k

∑ ∑ Y 2ij=?
i=1 j=1

We know that

Y ij =μ+ α j + β i +∈ij
2 2
Y ij ¿ ( μ+α j + β i +∈ij )
2 2 2
Y ij =μ +(α ¿¿ j) +¿ ¿
r k r k

∑ ∑ Y =∑ ∑ ¿ ¿ ¿ 2
ij
i=1 j=1 i=1 j=1

[∑ ∑ ]
r k k r
E Y =rk μ +r ∑ (α ¿¿ j) +k ∑ (β¿¿ i)2 ¿ ¿
2
ij
2 2

i=1 j=1 j=1 i=1

r k
E[Total SS] =E[ ∑ ∑ Y ij ¿−E[CF]
2

i=1 j=1

k r
E[Total SS] = r ∑ ( α ¿¿ j) +k ∑ ( β ¿¿ i) +(rk −1)σ ¿ ¿
2 2 2

j=1 i=1
Treatment Sum of Squares

E[Treatment SS]
k

∑ Y . j2
Treatment SS= j=1 −CF
r
r r
Y . j=∑ Y ij=∑ (μ+ α j+ βi +∈ij )
i=1 i=1

r r
Y . j=rμ +rα j + ∑ β i + ∑ ∈ij
i=1 i=1

As for fixed effect models

r

∑ β i=0
j=1

So,
r
Y . j=rμ +rα j + ∑ ∈ij
i=1

Squaring both side

2
Y . j =¿
2 2 2 2 2
Y . j =r μ +r α j+ ¿ ¿
Apply summation on both sides
k k

∑ Y . j 2= ∑ ¿ ¿
j=1 j=1

k k

∑ Y . j =r 2 2 2
k μ +r 2
∑ α 2 j +¿
j=1 j=1

Divide by r on both sides

k

∑ Y . j2 1
j =1
= ¿
r r
 k

∑ Y . j2 1
k
j =1
=rk μ + r ∑ α 2 j+ ¿
2
r j=1 r
As we know that

¿,
So
k

∑ Y . j2 1
k
2
❑ ❑ r k k r k k r k
j =1
=rk μ + r ∑ α j+ ( ∑ ∑ ∈ij + ∑ ∑ ∈ij ∈gh)+2 rμ ∑ α j +2 μ ∑ ∑ ∈ij +2 ∑ α j ∑ ∑ ∈ij
2 2
r j=1 r i=1 j=1 i ≠ g j≠ h j=1 i=1 j=1 j=1 i=1 j=1

Now taking expectation on both sides

k

∑ Y . j2 1
k
2
❑ ❑ r k k r k k r k
E [ j=1 ]=E[rk μ2 +r ∑ α 2 j + ( ∑ ∑ ∈ij +∑ ∑ ∈ij ∈gh )+2 rμ ∑ α j +2 μ ∑ ∑ ∈ij +2 ∑ α j ∑ ∑ ∈ij ]
r j=1 r i=1 j=1 i ≠ g j ≠h j=1 i=1 j=1 j=1 i =1 j=1

∑ Y . j2 1
k
E[ j=1
]=rk μ +r ∑ α 2 j + (rk σ 2+ 0)+ 2rμ (0)+2 μ (0)+2(0)
2
r j =1 r
k

∑ Y . j2 k
E[ j=1
]=rk μ +r ∑ α 2 j +(k σ 2)
2
r j =1

E(Treatment SS)=E ¿
2 2
E(CF )=rk μ +σ
k
E ( Treatment SS )=rk μ 2+ r ∑ α 2 j+ ( k σ 2 ) −(rk μ 2+ σ 2)
j=1

k
E(Treatment SS)=r ∑ α j2+ ( k−1 ) σ 2
j=1

Block Sum of Squares

E[Block SS]
r

∑ Y i .2
Block SS= i=1 −CF
k
k k
Y i . =∑ Y ij =∑ (μ +α j + β i+ ∈ij )
j=1 j=1
 k k
Y i . =kμ+ ∑ α j +k β i+ ∑ ∈ij
j=1 j=1

As for fixed effect models

r

∑ α j=0
j=1

So,
k
Y i . =kμ+ k βi + ∑ ∈ij
j=1

Squaring both side

2
Y i . =¿
2 2 2 2 2
Y i . =k μ + k β i +¿ ¿
Apply summation on both sides
r r

∑ Y i. =∑ ¿ ¿ 2

i=1 i=1

r r r k r k r r k r
2
∑ Y i.2=r k 2 μ2 +k 2 ∑ β 2i +(∑ ∑ ∈ij) +2 k 2 μ ∑ βi +2 kμ ∑ ∑ ∈ij +2 k ∑ βi ∑ ∑ ∈ij
i=1 i=1 i=1 j=1 i =1 j=1 i=1 i=1 j=1 i=1

Divide by k, apply properties, an taking expectation on both sides

k

∑ Y . j2 1 2
r r k r k r r k r
j =1
= [r k 2 μ 2+ k 2 ∑ βi2+( ∑ ∑ ∈ij ) +2 k 2 μ ∑ β i +2 kμ ∑ ∑ ∈ij + 2 k ∑ βi ∑ ∑ ∈ij ]
k k i=1 i=1 j=1 i=1 j=1 i=1 i=1 j=1 i=1

As we know that

¿,
So
r

∑ Y i .2 r
E[ i=1
]=rk μ + k ∑ β2i +(r σ 2 )
2
r i =1

E(Block SS)=E ¿
 2 2
E(CF )=rk μ +σ
r
E ( Block SS )=rk μ + k ∑ β2i +(r σ 2 )−(rk μ 2+ σ 2 )
2

i=1

r
E(Block SS)=k ∑ β2i + ( r −1 ) σ 2
i=1
Error Sum of square

E(Error SS) = ?

Error SS = Total SS -Treatment SS – Block SS

E(Error SS) = E(Total SS) – E(Treatment SS) – E(Block SS)

k r
E[Total SS] = r ∑ ( α ¿¿ j) +k ∑ ( β ¿¿ i) +(rk −1)σ ¿ ¿
2 2 2

j=1 i=1

k
E(Treatment SS)=r ∑ α j2+ ( k−1 ) σ 2
j=1

r
E(Block SS)=k ∑ β2i + ( r −1 ) σ 2
i=1

( )
k r k r
E(Error SS) = r ∑ ( α ¿¿ j) +k ∑ ( β ¿¿ i) + ( rk −1 ) σ − r ∑ α j + ( k−1 ) σ −(k ∑ β i + ( r−1 ) σ )¿ ¿
2 2 2 2 2 2 2

j=1 i=1 j=1 i=1

E(Error SS) = ( rk −1 ) σ 2−( ( k−1 ) σ 2) −( ( r−1 ) σ 2 )

E(Error SS) = (r −1)(k−1)σ 2

E(MSE) = σ 2
 Latin square designs (TxT)
The major feature of Latin Square Designs is its capacity to handle two sources of variation
simultaneously among the treatments. Which are treated as two independent blocking criteria. These
blocks are known as row-block and column-block. This is called as double blocking. Both source of
variations are perpendicular to each other. These designs are applied in field trials, where the
experimental area has two fertility gradients running perpendicular to each other, or in the green house
experiments in which the experimental pots are arranged in straight line perpendicular to the sheets or
walls of the green house such that the difference among rows and the distance from the wall are
expected to be two major extrenous sources of variation, or in laboratory experiments where the trial
are replicated over time such that the difference among the experimental units conducted at the same
time and those conducted over different time period constitute the two known sources of variations

Rows of tree
Water A B C
Channel B C A
C A B

Randomization and layout plan

Suppose we have five treatments (A, B, C, D, E) for this we need 5x5 LS-Designs, which means we should
layout the experiment with five rows and five columns

A B C D E
B C D E A
C D E A B
D E A B C
E A B C D
First of all randomize row arrangement by using random numbers then randomize column arrangement
by using random numbers. Generate five random numbers on your calculator

Random numbers Sequence Rank

628 1 3
846 2 4
475 3 2
902 4 5
452 5 1

C D A E B
D E B A C
B A E C D
E C D B A
A B C D E
Generate random numbers for the columns

Random numbers Sequence Rank

792 1 4
032 2 1
947 3 5
293 4 3
196 5 2

Row Column Number

Number 1 2 3 4 5
1 E C B A D
2 A D C B E
3 C B D E A
4 B E A D C
5 D A E C b

Statistical analysis

For a statistical analysis ANOVA for LS-Designs is used given as follow

SOV df SS MS Fcal F tab/P-value

Rows r-1 = 4
Columns c-1 = 4
Treatments t-1 = 4
Error 12
Total rc-1 = 24
What is the effect of error degrees of freedom on thetrials?

Example: -

An experiment was conducted with three maize varieties and a check variety, the experiment was
laidout under LS-Designs, Analyse the data given below

C-1 C-2 C-3 C-4 Total

R-1 1640(B) 1210(D) 1425(C) 1345(A)
R-2 1475(C) 1185(A) 1400(D) 1290(B)
R-3 1670(A) 710(C) 1665(B) 1180(D)
R-4 1565(D) 1290(B) 1655(A) 660(C)
Total

A B C D
1670 1640 1475 1565
1185 1290 710 1210
1655 1665 1425 1400
1345 1290 660 1180
Find correction factor

CF = (GT)2/n
t r
Total SS=∑ ❑ ∑ y ij 2−CF
j=1 i=1

∑ T j2
Treatment SS= j=1 −CF
r
b

∑ Ri2
Rows SS= r=1 −CF
t
b

∑ C j2
Columns SS= r =1 −CF
t
Error SS = Total SS- Treatment SS – Rows SS – Columns SS

ANOVA

SOV df SS MS Fcal F tab (5%) F tab (1%)

Rows 3 30154.69 10051.56 0.465NS 4.7571 9.7795
Column
3 827342.19 275780.73 12.769** 4.7571 9.7795
s
Trt 3 426842.19 142280.73 6.588* 4.7571 9.7795
Error 6 129584.38 21597.40
Total 15 1413923.44
 Comparison Among Treatment
Means
When we conduct an experiment we have the following hypothesis

H0: All the treatment means are equal

H1: At least two treatment means are not equal

If we have three treatments T1, T2, and T3

T1 and T2 are not equal

T1 and T3 are not equal

T2 and T3 are not equal

There are two types of comparison among treatment means

1. Planned pair comparison

2. Unplanned pair comparison

In planned pair comparison a specific pair of treatments to be compared was identified before the start
of the experiment. For example a control is compared with other treatments

In unplanned pair comparison is that in which no specific comparison is chosen in advance, instead
every possible treatment means is compared to identify pairs of treatments that are significantly
different

For this there are two commonly used test named least significant difference test (LSD-Test) and
Duncan’s multiple range test (DMRT)

1. Least Significant difference test (LSD-Test)

The LSD-Test is a most commonly used procedure for making pair comparison. This procedure provides
a single LSD-Value, which serves as a boundary between significant and nonsignificant difference among
the treatment pairs for a given level of significance such as if the difference between pair means exceeds
LSD-Value then the difference is significant otherwise it is nosignificant.
The procedure for applying LSD_Test is as follow

1. first of all find the means of all treatments

2. Arrange these in ascending or descending order of magnitude

N2 N2 N3 N4 N5
85.7 94.2 97.4 101.5 96.2

N4 N3 N5 N2 N2
101.5 97.4 96.2 94.2 85.7

3. find the difference of each pair as follow

As there are five treatments so we have 10 pairs for comparison

Mean pairs Difference

N4-N2 101.5-85.7 15.8* >2.73
N4-N2 101.5-94.2 7.3* >2.73
N4-N5 101.5-96.2 5.3* >2.73
N4-N3 101.5-97.4 4.1* >2.73
N3-N2 97.4-85.7 11.7* >2.73
N3-N2 97.4-94.2 3.2* >2.73
N3-N5 97.4-96.2 1.2NS <2.73
N5-N2 96.2-85.7 10.5* >2.73
N5-N2 96.2-94.2 2.0 NS <2.73
N2-N2 94.2-85.7 8.5* >2.73

N4 N3 N5 N2 N2
101.5 97.4 96.2 94.2 85.7
a b bc c d

4. compute the LSD-Value as follow

LSD α =(t α )( sd ),

Where t α is table value of t and,

sd , is standard error of difference between means and it is computed as follow

sd =
√ 2 MSE ,
r
sd =
√
2(3.398) s =1.5051
3
, d ,

t 0.05 ,10=1.8125 ,
LSD 0.05=( 1.8125 )( 1.5051 )=2.733.

N2 N2 N3 N4 N5
R1 88.4 92.4 96.8 101.2 97.3
R2 83.4 96.7 97.1 99.8 96.5
R3 85.3 93.6 98.3 103.4 94.7
Total 257.1 282.7 292.2 304.4 288.5
Averag
e 85.7 94.2 97.4 101.5 96.2

ANOVA
SOV SS df MS F P-value F crit
Treatments 408.0493 4 102.0123 30.02** 0.000 3.4781
Error 33.98 10 3.398
Total 442.0293 14

Calculate Grand Mean and coefficient of variation as follow

G 1424.9
Grand Mean= = = 94.99,
n 15

C.V=
σ
x 100=
√ MSE X 100= √3.40 X 100=1.94 %,
x Grand Mean 94.99
C.V tells about the degree of precision in the experiment. 1.94% C.V tells about the best index of
reliability in the experiment. Which shows that the experimental error is very low. Higher the C.V lower
the reliability/precision. In general for lab experiments, green house experiments pot experiments or
where the experimental material is homogeneous and the environmental variations can be controlled
the range of C.V should upto 5%, in the field experiments the C.V should be 5% to 10 %. Whaich are
good degree of reliability. As the C.V increases the reliability of the experiment decreases.
 Duncan’s Multiple Range Test
(DMRT)
As the number of treatments increase the no of pairs of treatments also increase, if there are six

treatments then we have (62)=15, no of pairs. If there are 10 treatments then we have (102)=45, no of
pairs and if we we have 15 treatments then there are ( )=105 , no of pairs. As the treatment
15
2
combinations increase the precision decreases and the level of significane also increase as the difference
between the largest and smallest value increased by 29% for five treatments, 63 % for 10 treatments
and 83 % for 15 treatments. So to overcome this we use DMRT which gives diffent LSD-values for each
set of difference.

The procedure for DMRT is same as LSD-Test as follow

1. Arrange the treatment means in ascending or descending order

2. take the difference between all possible pairs/treatment combinations

3. compute the LSD-value as follow

(r p )(s d )
Rp= , for p = 2,3,4,…
√2
Where t α is table value of t and,

sd , is standard error of difference between means and it is computed as follow

sd =
√ 2 MSE ,
r
sd =
√
2(3.398) s =1.5051
3
, d ,
And, known as

r p, is the table value for Duncan’s known as significant studentized range obtained from appendix B
page 542 and 543 of the book Introduction to Statistical Theory part-II by Sher Muhammad Chaudhary

At 5% SL

P 2 3 4 5
rp 3.15 3.30 3.37 3.43
(r p )(sd ) 4.74 4.97 5.07 5.16
(r p )( sd )
√2 3.35 3.51 3.59 3.65

Mean pairs Difference

N4-N2 101.5-85.7 15.8* >3.35
N4-N2 101.5-94.2 7.3* >3.35
N4-N5 101.5-96.2 5.3* >3.35
N4-N3 101.5-97.4 4.1* >3.35
N3-N2 97.4-85.7 11.7* >3.51
N3-N2 97.4-94.2 3.2 NS <3.51
N3-N5 97.4-96.2 1.2NS <3.51
N5-N2 96.2-85.7 10.5* >3.59
N5-N2 96.2-94.2 2.0 NS <3.59
N2-N2 94.2-85.7 8.5* >3.65
At 1% SL

P 2 3 4 5
rp 4.48 4.73 4.88 4.96
(r p )(sd ) 6.74 7.12 7.34 7.47
(r p )( sd )
√2 4.77 5.03 5.19 5.28
Mean pairs Difference 5% 1%
N4-N2 101.5-85.7 15.8** >3.35 >4.77
N4-N2 101.5-94.2 7.3** >3.35 >4.77
N4-N5 101.5-96.2 5.3** >3.35 >4.77
N4-N3 101.5-97.4 4.1* >3.35 <4.77
N3-N2 97.4-85.7 11.7** >3.51 >5.03
N3-N2 97.4-94.2 3.2 NS <3.51 <5.03
N3-N5 97.4-96.2 1.2NS <3.51 <5.03
N5-N2 96.2-85.7 10.5** >3.59 >5.19
N5-N2 96.2-94.2 2.0 NS <3.59 <5.19
N2-N2 94.2-85.7 8.5** >3.65 >5.28

N4 N3 N5 N2 N2
101.5 97.4 96.2 94.2 85.7
a bc bc c d
 N2 N2 N3 N4 N5 Total
R1 88.4 92.4 96.8 101.2 97.3
R2 83.4 96.7 97.1 99.8 96.5
R3 85.3 93.6 98.3 103.4 94.7
Total 257.1 282.7 292.2 304.4 288.5
Average 85.7 94.2 97.4 101.5 96.2

ANOVA (RCBD)
SOV SS df MS F P-value F crit
0.70933 0.35466
Block 2 0.08528NS 0.9191
3 7
408.049 102.012 24.52907*
Treatments 4 0.0002
3 3 *
33.2706
Error 8 4.159
7
442.029 1
Total
3 4

ANOVA (CRD)
SOV SS df MS F P-value F crit
Treatments 408.0493 4 102.0123 30.02** 0.000 3.4781
Error 33.98 10 3.398
Total 442.0293 14

Calculate Grand Mean and coefficient of variation as follow

G 1424.9
Grand Mean= = = 94.99,
n 15

x 100= √ X 100= √
σ MSE 4.159
C.V= X 100=2.15 %,
x Grand Mean 94.99
Relative Efficiency Of Randomized Complete Block Designs Over Completely Randomized Designs

We examine the block efficiency

For this purpose the relative efficiency is given as

( r−1 )∗MS blocks+ r ( t−1 )∗MSE

ℜ= ,
( rt−1 )∗MSE
(3−1 )∗0.3547+3 ( 5−1 )∗4.159
ℜ= ,
( 3∗5−1 )∗4.159
 ( 2 )∗0.3547+ 12∗4.159
ℜ= ,
14∗4.159
ℜ=0.869=86.9 %,
If the error degrees of freedom are less than 20 then multiply RE with the factor “k” where is given as

k=
[ ( r−1 ) ( t−1 )+ 1 ] [t ( r−1 ) +3] = [ ( 2 ) ( 4 ) +1 ] [5 (2 )+3 ] = [ 9 ] [13] = 117 =0.9669
[ ( r−1 ) ( t−1 )+ 3 ] [t ( r−1 )+1] [ ( 2 ) ( 4 ) +3 ] [5 ( 2 )+ 1] [ 11 ] [11] 121
Adjustef RE = RE * k = 0.869 * 0.9669 = 0.841 = 84.1

Relative efficiency of LS-Designs over CR-Designs and RCB-Designs

SOV df SS MS Fcal F tab (5%) F tab (1%)

Rows 3 30154.69 10051.56 0.465NS 4.7571 9.7795
Column
3 827342.19 275780.73 12.769** 4.7571 9.7795
s
Trt 3 426842.19 142280.73 6.588* 4.7571 9.7795
Error 6 129584.38 21597.40
Total 15 1413923.44
The relative efficiency of LS-Designs over CR-Designs is given as

MS Rows+ MS Coloumns+ ( t−1 ) MS Error

ℜ ( CRD )= ,
( t+1 ) MS Error
10051.56+275780.73+ ( 4−1 ) 21597.40 350624.5
ℜ ( CRD )= = =3.25=325 %,
( 4+1 ) 21597.40 107987
The LS-Designs are 225% more efficient than CR-Designs, it also shows that we need 2.25 more
replications to achieve the required efficiency

Now the relative efficiency of LS-Designs as compared to RCB-Designs with row blocking and coloumn
blocking

MS Rows+ ( t−1 ) MS Error 10051.56+ ( 4−1 ) 21597.40

ℜ ( RCBD , rows )= = =0.867
( t ) MS Error ( 4 ) 21597.40
MS Coloumns+ ( t−1 ) MS Error 275780.73+ ( 4−1 ) 21597.40
ℜ ( RCBD , col )= = =3.94 ,
( t ) MS Error ( 4 ) 21597.40
If the degrees of freedom for error are less than 20 then adjust RE by multiplying with the factor “k”

Where “k” is given as

k=
[ (t−1 )( t−2 ) +1 ] [ ( t−1 )2 +3] = [ ( 4−1 ) ( 4−2 ) +1 ] [ ( 4−1 )2 +3] = [ ( 3 ) ( 2 ) +1 ] [9+3]
,
[ (t−1 )( t−2 ) +3 ] [( t−1 )2 +1] [ ( 4−1 ) ( 4−2 ) +3 ][ ( 4−1 )2+1] [ ( 3 ) ( 2 ) +3 ] [9+1]
[ 7 ] [12] 84
k= = = 0.93
[ 9 ] [10] 90
275780.73+ ( 4−1 ) 21597.40
ℜ ( RCBD , col , Adjusted )=3.94∗0.93=3.66 = =3.94
( 4 ) 21597.40
 Latin Square Designs
Random Effect Models

Mean square error for LS-Designs

The model for LS-Designs is given as

y ij(k) =μ +α i+ β j+ γ k + ∈ij (k ), where i=j=k=1,2,3,…,p

μ=overall mean ,
α i=row effect ,

β j =coloumn effect ,

γ k =treatment effect ,

∈ij(k)=randomerror NII ¿ ,

The assumptions are

2 2
E(∈¿¿ ij (k ))=0 ¿, E(∈ij (k ))❑=σ , E(∈ ¿ ¿ ij (k )∈ gh(n))=0 ¿ for i ≠ g, j ≠ h, k ≠ n

In random effect models α i , β j , γ k , are random with the assumptions

' 2
α i s NII (0 , σ α )

E(α ¿¿ i)=0 ¿, E(α 2i )❑ =σ 2α, E(α ¿ ¿ i α g )=0 ¿, i ≠ g

' 2
β i s NII (0 , σ β ),

E( β¿¿ j)=0 ¿, E( β2j )❑ =σ 2β , E( β ¿ ¿ j β h )=0 ¿, j ≠ h,

' 2
γ i s NII (0 , σ γ ),

E(γ ¿ ¿ k )=0 ¿, E(γ 2k )❑=σ 2γ , E(γ ¿ ¿ k γ n)=0¿ , k ≠ n

α i , β j , γ k , are independent

E [ ( α i ) ( β j ) ]=0, E [ ( α i ) ( γ k ) ] =0,

E [ ( β j )( γ k ) ]=0 ,

Correction factor
2
G
CF= 2,
p
 p p
G=∑ ∑ (μ+ α i + β j + γ k +∈ij (k )),
j=1 i=1

p p p p p
G= p 2 μ+ p ∑ α i+ p ∑ β j+ p ∑ γ k + ∑ ∑ (∈ij (k )),
i=1 j=1 k=1 j=1 i=1

Squaring on both sides

p p p p p
2
G =( p μ+ p ∑ α i + p ∑ β j + p ∑ γ k + ∑ ∑ ( ∈ij (k )) ) ,
2 2

i=1 j =1 k=1 j=1 i=1

( ) ( ) (
p p p p p p p p
G2= p 4 μ 2+ p2 ( ∑ α i ) + p 2 (∑ β j ) + p2 ( ∑ γ k ) +( ∑ ∑ ( ∈ij ( k ) ) ) + ( p2 μ ) p ∑ α i +2 ( p2 μ ) p ∑ β j + 2 ( p 2 μ ) p ∑ γ
2 2 2 2

i =1 i=1 i =1 j=1 i=1 i=1 j=1 k =1

We know that
p p ❑
2
( ∑ α i ) =∑ α 2i + ∑ α i α g,
i=1 i=1 i≠ g

p p ❑
2
( ∑ β j) =∑ β + ∑ β j β h, 2
j
i=1 j=1 j ≠h

p p ❑
2
( ∑ γ k ) =∑ γ 2k + ∑ γ k γ n,
i=1 i=1 k≠ n

p p p p p p
2
( ∑ ∑ ( ∈ij ( k ) ) ) =∑ ∑ ( ∈2ij (k ) ) + ∑ ∑ ∈(¿ ij(k) ∈ gh(n))¿,
j=1 i=1 j=1 i=1 i=1 j=1

Dividing by p2, and taking expectation on both sides, we get

2
G
E(CF )=E[ ]
2 ,
p

( ) ( )
p p p p p p p
1
E [ p 4 μ 2+ p 2( ∑ α i) + p 2 ( ∑ β j ) + p 2 (∑ γ k ) +(∑ ∑ ( ∈ij (k )) ) + ( p2 μ ) p ∑ α i +2 ( p2 μ ) p ∑ β j +2 ( p 2
2 2 2 2
E ( CF ) = 2
p i=1 i=1 i=1 j=1 i=1 i=1 j=1

1
E ( CF ) = 2
E ¿,
p
1 4 2 2
[ p μ + p [ pσ α +0]+ p [ pσ β +0 ]+ p [ pσ γ +0 ]+ p [ σ e + 0 ] +0 ],
2 2 2 2 2 2 2
E ( CF ) = 2
p
2 2 2 2 2 2
E ( CF ) =p μ + pσ α + pσ β + pσ γ + σ e ,
 [ ]
n n n
E ∑ x i =∑ E[ x¿ ¿i¿]=∑ μ=nμ ¿ ¿,
i=1 i=1 i=1

[∑ ] ∑
p p p
E[α i ]=∑ σ α = p σ α ,
2 2 2 2
E αi =
i=1 i=1 i=1

[∑ ] ∑
p p p
E[ β j ]=∑ σ β = p σ β,
2 2 2 2
E βj =
j=1 i=1 i=1

Hence,
2 2 2 2 2 2
E ( CF ) =p μ + pσ α + pσ β + pσ γ + σ e ,
Total Sum of Square
p p
Total SS=∑ ∑ y 2ij(k )−CF ,
i=1 j=1

The model for LS-Designs is given as

y ij(k) =μ +α i+ β j+ γ k + ∈ij (k ),

y 2ij(k) =(μ+ α i + β j + γ k +∈ij(k) )2,

[ ]
p p
E ∑ ∑ y 2ij( k ) = p 2 μ2 + p2 σ 2α + p 2 σ 2β + p 2 σ 2γ + p 2 σ 2e,
i=1 j=1

As
2 2 2 2 2 2
E ( CF ) =p μ + pσ α + pσ β + pσ γ + σ e
2 2 2 2
E [Total SS ]= p ( p−1 ) σ α + p ( p−1 ) σ β + p ( p−1 ) σ γ +( p−1)( p+1)σ e ,

Treatment SS
p

∑ T 2k ,
Treatment SS= k=1 −CF
p
p
T k =∑ y ij(k ),
i= j

❑
T k =∑ (μ+α i+ β j +γ k +∈ij(k)) ,
i= j

p p p
T k = p μ+ ∑ α i+ ∑ β j + pγ k + ∑ ∈ij(k)
i=1 j =1 i= j
Row SS
p

∑ R2j ,
Row SS= j=1 −CF
p
p
R j=∑ y ij(k) ,
j=1

p
R j=∑ ¿ ¿ ,
j=1

2 2
E ( Row SS )= p ( p−1 ) σ α +(p−1)σ
Coloumn SS
p

∑ C2j ,
Column SS= i=1 −CF
p
p
C i=∑ y ij(k),
I=1

p
C i=∑ ¿ ¿
I=1

2 2
E ( Column SS )= p ( p−1 ) σ β +( p−1)σ

Error SS

E ( Error SS )=E [ Total SS ] −E ( Column SS )−E ( Row SS )−E ( Treatment SS ) ,

2
E ( Error SS )=Mean Square Error ( MSE ) =σ

Latin Square Designs

Fixed Effect Models

Mean square error for LS-Designs

The model for LS-Designs is given as

y ij(k) =μ +α i+ β j+ γ k + ∈ij (k ), where i=j=k=1,2,3,…,p

μ=overall mean ,
α i=row effect ,

β j =coloumn effect ,

γ k =treatment effect ,

∈ij(k)=randomerror NII (0 , σ 2 ),

The assumptions are

2 2
E(∈¿¿ ij (k ))=0 ¿, E(∈ij (k ))❑=σ , E(∈ ¿ ¿ ij (k )∈ gh(n))=0 ¿ for i ≠ g, j ≠ h, k ≠ n

In fixed effect models α i , β j , γ k , are fixed with the assumptions

' 2
α i s NII (0 , σ α )

E(α¿¿ i)=0 ¿, E(α 2i )❑=σ 2α, E(α ¿ ¿ i α g )=0 ¿, i ≠ g

 ' 2
β i s NII (0 , σ β ),

E( β¿¿ j)=0 ¿, E( β2j )❑ =σ 2β , E( β ¿ ¿ j β h )=0 ¿, j ≠ h,

' 2
γ i s NII (0 , σ γ ),

E(γ ¿ ¿ k )=0 ¿, E(γ 2k )❑=σ 2γ , E(γ ¿ ¿ k γ n)=0¿ , k ≠ n

α i , β j , γ k , are independent

E [ ( α i ) ( β j ) ]=0, E [ ( α i ) ( γ k ) ] =0,

E [ ( β j )( γ k ) ]=0 ,
 Least square estimates for CR-Designs, RCB-Designs, and LS-Designs

Completely Randomized Designs

The basic Statistical model for CR-Designs is

Y ij =μ+ α j +∈ij ,

Y^ ij =^μ + α^ j , where ^μ∧ α^ j, are the estimates for ¿ μand α j,

We know that
r k
S=∑ ∑ (Y ij −Y^ ij )2,
i=1 j=1

r k
S=∑ ∑ (Y ij − μ^ −α^ j )2,
i=1 j=1

Now differentiate S partially with respect to ^μ

r k
∂S
=∑ ∑ (Y ij− μ^ −α^ j )2,
∂ μ^ i=1 j=1
r k
∂S
=2 ∑ ∑ ( Y ij −^μ− α^ j ) (−1),
∂ μ^ i =1 j=1

∂S
=−2 ¿,
∂ μ^
∂S
Now put =0, so
∂ μ^
−2 ¿,
r k k

∑ ∑ ( Y ij) =rk ^μ +r ∑ α^ j , …………………………………….. (A)

i=1 j=1 j=1

Now for treatments

As there are k treatments so,

r r r r
S=∑ (Y ij − μ^ −α^ 1)2 +∑ (Y ij− μ^ −α^ 2)2 + ∑ (Y ij −^μ− α^ 3 )2+ …+ ∑ (Y ij −^μ −α^ k )2,
i=1 i=1 i=1 i=1

^ 1, α^ 2, …, α^ k ,
Now differentiate S partially with respect to α
r
∂S
=−2 ∑ (Y ij −^μ −α^ 1) ,
∂ α^ 1 i=1
 r
∂S
=−2 ∑ (Y ij −^μ −α^ 2) ,
∂ α^ 2 i=1

.
r
∂S
=−2 ∑ (Y ij −^μ− α^ k ),
∂ α^ k i=1

∂S
=0,
∂ α^ j
r
−2 ∑ ( Y ij −^μ −α^ j )=0,
i=1

∑ ( Y ij−^μ −α^ j ) =0,

i=1

Apply summation
r

∑ ( Y ij) −r ^μ−r α^ j =0,

i=1

∑ ( Y ij) =r ^μ +r α^ j, …………………………………………………… (B)

i=1

Y . j=r ^μ +r α^ j
k
As we know that ∑ α
^ j=0 , so equation (A) becomes
j=1

r k

∑ ∑ ( Y ij) =rk ^μ,

i=1 j=1

Y .. ¿
^μ= = =Y ..,
rk rk
Now from equation (B)

Y . j=r ^μ +r α^ j ,

Divide r on both sides

Y.j
= μ^ + α^ j,
r
Put the value of ^μ
Y.j
=Y ..+ α^ j,
r
Y.j
α^ j = −Y ..,
r
α^ j =Y . j−Y ..,

^ j in Completely Randomized Designs are

So, the estimates of ^μ and α

^μ=Y ..,

α^ j =Y . j−Y ..
 Randomized Complete Block Designs

The basic Statistical model for RCB-Designs is

Y ij =μ+ α j + β i +∈ij,

The estimated model for RCB-Designs is given as

Y^ ij = ^μ + α^ j + β^ i, where ^μ∧ α^ j, are the estimates for ¿ μand α j,

We know that
r k
S=∑ ∑ (Y ij −Y^ ij )2,
i=1 j=1

r k
S=∑ ∑ (Y ij − μ^ −α^ j− ^βi )2,
i=1 j=1

^ j,
To find estimates for unknown parameters μ, α j , and β i, differentiate S partially with respect to ^μ, α
and ^β iand equate them equal to zero
r k
∂S
=∑ ∑ (Y ij− μ^ −α^ j − ^βi )2=0 ,
∂ μ^ i=1 j=1
r k

∑ ∑ (Y ij−^μ −α^ j− β^ i )=0,

i=1 j=1

Applying the summation we get

r k k r

∑ ∑ Y ij−rk ^μ−r ∑ α^ j−k ∑ ^β i=0, ……………………………… (A)

i=1 j=1 j=1 i=1

Now for blocks

k k k
S=∑ ( Y ij −^μ −α^ j− β^ 1 ) + ∑ (Y ij −^μ− α^ j− ^β 2)2 +…+ ∑ (Y ij −^μ− α^ j− ^β r )2
2

j=1 j=1 j=1

Now differentiate S partially with respect to ^β 1, α

^ 2, …, ^β r ,
k
∂S
=−2 ∑ (Y ij −^μ −α^ j − ^β1 ),
∂ ^β1 j=1

k
∂S
=−2 ∑ (Y ij −^μ −α^ j − ^β 2),
∂ ^β2 j=1

.
 k
∂S
=−2 ∑ (Y ij −^μ −α^ j− β^ r ),
∂ ^βr j=1

The general term is

k
∂S
=−2 ∑ (Y ij −^μ− α^ j − ^β i),
^
∂ βi j=1

put

∂S
=0 ,
∂ ^βi
k
−2 ∑ (Y ij− μ^ −α^ j− ^βi )=0 ,
j=1

∑ (Y ij−^μ− α^ j− ^β i)=0,
j=1

Applying summation
k k

∑ Y ij−k μ^ −∑ α^ j −k ^β i=0,……………………….. (B)

j=1 j=1

For treatments
r r r
S=∑ ( Y ij −^μ −α^ 1 − ^βi ) + ∑ (Y ij − μ^ −α^ 2− ^β i)2 +…+ ∑ (Y ij −^μ− α^ k − ^β i)2
2

i=1 i=1 i=1

^ 1, α^ 2, …, α^ k ,
Now differentiate S partially with respect to α
r
∂S
=−2 ∑ (Y ij −^μ −α^ 1− ^β i),
∂ α^ 1 i=1

r
∂S
=−2 ∑ (Y ij −^μ −α^ 2− β^ i) ,
∂ α^ 2 i=1

.
k
∂S
=−2 ∑ (Y ij −^μ− α^ k − ^βi ),
∂ α^ k j=1

The general term is

 r
∂S
=−2 ∑ (Y ij − μ^ −α^ j− β^ i ),
∂ α^ j i=1

put

∂S
=0,
∂ α^ j
r
−2 ∑ (Y ij− μ^ −α^ j− ^βi )=0 ,
i=1

∑ (Y ij−^μ− α^ j− ^β i)=0,
i=1

Applying summation
r r

∑ Y ij−r μ^ −r α^ j−∑ β^ i=0,……………………….. (C)

i=1 i=1

Equations (A), (B), and (C) can be written as

r k k r

∑ ∑ Y ij=rk ^μ +r ∑ α^ j +k ∑ β^ i, ……………………………… (A)

i=1 j=1 j=1 i=1

k k

∑ Y ij=k μ^ +∑ α^ j +k ^β i,……………………….. (B)

j=1 j=1

r r

∑ Y ij=r μ^ + r α^ j+∑ ^βi ,……………………….. (C)

i=1 i=1

Treatment1 Treatment2 Treatment3

Rep1 Y11 Y12 Y13 Y1.
Rep2 Y21 Y22 Y23 Y2.
Rep3 Y31 Y32 Y33 Y3.
Y.1 Y.2 Y.3 Y..
I = 1 to 3

J = 1 to 3
k r
use the assumption ∑ α
^ j=0∧∑ ^β i=0, so equation (A) becomes
j=1 i=1

r k

∑ ∑ Y ij=rk ^μ,
i=1 j=1
Y ..=rk μ^ ,

Y ..
^μ= ,
rk
^μ=Y .., N

Now equation (B)

k

∑ Y ij=k μ^ +k ^β i,
j=1

Y i . =k ^μ +k ^β i,

Y i.
^β = − ^μ,
i
k
^β =Y −Y ,
i i. ..

Now equation (C)

r

∑ Y ij=r μ^ + r α^ j,
i=1

α^ j =Y . j−Y ..

Missing observations in RCB-

Designs
Case-I: one missing observation

Y ij =μ+ α j + β i +∈ij,

Where i = 1, 2, …, r

And j = 1, 2, …, k

The estimated model for RCB-Designs is given as

Y^ ij = ^μ + α^ j + β^ i,
^ id
Let the missing observation is in the ith block and jth treatment and is denoted by Y

Treatment(1 Treatment(2) Treatment(3) Treatment(j) Treatment(k Total

) )
Blo(1) Y11 Y12 Y13 Y1j Y1k Y1.
Blo(2) Y21 Y22 Y23 Y2j Y2k Y2.
Blo(3) Y31 Y32 Y33 Y3j Y3k Y3.
Blo(i) Yi1 Yi2 Yi3 Yij Yik B́c + Y^ id
Blo(r) Yr1 Yr2 Yr3 Yrj Yrk Yr.
Total Y.1 Y.2 Y.3 T´ d + Y^ id Y.k Ǵ+ Y^ id
The total no of observations are (rk-1)

So the sums are

Grand Total is G=Ǵ+ Y^ id,

^ id,
Ith block Total is Bc = B́c + Y

^ id
Jth treatment total is T d=T´ d + Y

Treatment(1) Treatment(2) Treatment(3) Treatment(j)

Blo(1) Y11 Y12 Y13 Y11+Y12+Y13
Blo(2) Y21 Y22 A Y21+Y22+A
Blo(3) Y31 Y32 Y33 Y31+Y32+Y33
Blo(i) Y11+Y21+Y31 Y12+Y22+Y3 Y13+Y33+A
2
We also know that
Error SS = Total SS-Block SS-Treatment SS
2
G
CF =
rk
r k
Total SS = ∑ ∑ Y ij −CF
2

i=1 j=1

∑ Y i .2
Block SS= i=1 −CF
k
k

∑ Y . j2
Treatment SS= j=1 −CF
r
After one missing observation in ith block and jth treatment

( Ǵ+ Y^ id )
2
CF=
rk
r ´k
Total SS=∑ ∑ Ý ij + Y^ id −CF ,
2 2

i=1 j=1

r
1
Block SS= [ ∑ B 2i + ( B́c + Y^ id ) ]−CF ,
2

k i=1
k
1
Treatment SS= [ ∑ T 2j + ( T´ d + Y^ id ) ]−CF ,
2

r j=1

[ ] [∑ ]
r ´k r k
1 1
Error SS=∑ ∑ Ý ij + Y^ id − ∑ Bi + ( B́c + Y^ id ) − T j + ( T´ d + Y^ id ) +CF , ….(A)
2 2 2 2 2 2

i=1 j=1 k i=1 r j=1

To estimate the missing observation apply method of least square and take the partial derivative of
^ ) and euate it to zero
equation (A) with respect to the missing observation (i.e. Y id

[ ] [ ]
r ´k r k
( Ǵ+ Y^ id )
2
∂ ESS ∂ 1
[ ∑ ∑ Ý 2ij + Y^ 2id − ∑ B + ( B́c + Y^ id ) − 1r ∑ T +(T´ d +Y^ id ) +
2 2 2 2
= ],
∂ Y^ id ∂ Y^ id i=1 j=1
i j
k i=1 j=1 rk
Case II: - two missing observations

There are three situations for two missing observations in RCB-Designs

i) in different blocks and different treatments

ii) in same block but for different treatments

iii) in different blocks but same treatments

now we proceed for first situation i.e. in different blocks and different treatments

Y ij =μ+ α j + β i +∈ij,

Where i = 1, 2, …, r

And j = 1, 2, …, k

The estimated model for RCB-Designs is given as

Y^ ij = ^μ + α^ j + β^ i,

Let the two missing observation are in denoted by M ∧N , suppose they are in block 3 treatment 2 and
block 4 treatment 5, respectively

Treatment(1 Treatment(2) Treatment(3) Treatment(j) Treatment(k Total

) )
Blo(1) Y11 Y12 Y13 Y1j Y1k Y1.
Blo(2) Y21 Y22 Y23 Y2j Y2k Y2.
Blo(3) Y31 N Y33 Y3j Y3k
Blo(i) Yi1 Yi2 Yi3 M Yik
Blo(r) Yr1 Yr2 Yr3 Yrj Yrk Yr.
Total Y.1 Y.3 Y.k Ǵ+ Y^ id
The total no of observations are (rk-1)

So the sums are

Grand Total is G=Ǵ+ M + N ,

block Total is BM =B´M + M ,

block Total is BN = B́ N + N ,

treatment total is T M =T´M + M

treatment total is T N =T´N + N

Treatment(1) Treatment(2) Treatment(3) Treatment(j)

Blo(1) Y11 B Y13 Y11+B+Y13
Blo(2) Y21 Y22 A Y21+Y22+A
 Blo(3) Y31 Y32 Y33 Y31+Y32+Y33
Blo(i) Y11+Y21+Y31 B+Y22+Y32 Y13+A+Y33 Ǵ+ A+ B
We also know that

Error SS = Total SS-Block SS-Treatment SS

2
G
CF=
rk
r k
Total SS = ∑ ∑ Y ij −CF
2

i=1 j=1

∑ Y i .2
Block SS= i=1 −CF
k
k

∑ Y . j2
Treatment SS= j=1 −CF
r
After two missing observations we have,
2
( Ǵ+ M + N )
CF=
rk
r ´k
Total SS=∑ ∑ Ý ij + M + N −CF ,
2 2 2

i=1 j=1

r
1
Block SS= [ ∑ B 2i + ( B´M + M ) + ( B´N + N ) ]−CF ,
2 2

k i=1
k
1
Treatment SS= [ ∑ T 2j + ( T´M + M ) + ( T´N + N ) ]−CF ,
2 2

r j=1
r ´k
1
Error SS=∑ ∑ Ý ij + M + N − ¿ , ….(A)
2 2 2

i=1 j=1 k
To estimate the missing observation apply method of least square and take the partial derivative of
equation (A) with respect to the missing observation (i.e. M ∧N ) respectively and euate it to zero
r ´k r k
( Ǵ+ M + N)
∂ ESS ∂ 1 1
[ ∑ ∑ Ý ij + M + N − [ ∑ Bi + ( B´M + M ) + ( B́N + N ) ]− [ ∑ T j + ( T´M + M ) + ( T´N + N ) ]+
2 2 2 2 2 2 2 2 2
=
∂M ∂ Y^ id i=1 j=1 k i=1 r j=1 rk
,,

∂ ESS 1 1 1
=0+2 M + 0− [ 0+2 ( B´M + M ) ( 1 ) +0 ]− [ 0+ 2 ( T´M + M ) ( 1 )+ 0 ] + ¿,
∂M k r rk
∂ ESS 2 2 2
=2 M − [ ( B´M + M ) ] − [ ( T´M + M ) ] + ¿,
∂M k r rk
∂ ESS
Put =0, so
∂M
2 ´ 2 2
2 M−
k
[ ( B M + M ) ] − [ ( T´M + M ) ] + ¿,
r rk
rk
Multiply by on both sides
2

rkM −r [ ( B´M + M ) ]−k [ ( T´M + M ) ] +¿ ,

rkM −rM−kN + M −r B´M −k T´M + Ǵ+ N=0,

( r −1 )( k−1 ) M + N =+r B´M + k T´M −Ǵ, ………………… (1)

Now differtiating with respect to N we get

r ´k r k
( Ǵ+ M + N)
2
∂ ESS ∂ 1 1
[ ∑ ∑ Ý + M + N − [ ∑ Bi + ( B M + M ) + ( B́N + N ) ]− [ ∑ T j + ( T M + M ) + ( T N + N ) ]+
´ ´ ´
2 2 2 2 2 2 2 2 2
=
∂N ∂ N i=1 j=1 ij k i=1 r j=1 rk
,,

∂ ESS 2 2 2
=2 N − [ ( B´N + N ) ] − [ ( T´N + N ) ] + ¿,
∂N k r rk
Equate it to zero

2 2 2
2 N−
k
[ ( B́N + N ) ] − [ ( T´N + N ) ] + ¿,
r rk
rk
Multiply by on both sides
2

rkN −r [ ( B´N + N ) ]−k [ ( T´N + N ) ] +¿,

(r −1)(k−1)N + M =r B´N + k T´ N −Ǵ , …(2),

By solving equation (1) and (2) simultaneously, we get

( r−1 )( k −1 ) ( r B´M +k T´M − Ǵ )−(r B´N + k T´N −Ǵ)

M= ¿
2
¿,
((r−1)(k −1)) −1

( r−1 )( k −1 ) ( r B´N + k T´N −Ǵ ) −(r B´M +k T´M −Ǵ)

M= ¿
2
¿,
((r−1)(k −1)) −1
 Incomplete Block Designs
When the block size is less than the number of treatments to be tested is known as incomplete block
designs. These designs were introduced by Yates to eliminate the heterogeneity when the number of
treatments becomes very large. As we know that precision of the estimate of the treatment effects
depends on the number of replication, similar is the case for the precision of the estimatesof difference
between two treatment effects. So when we have set of treatments is large than the block size we use
incomplete block designs. For example we have to test the quality of six tyres, on a given car only 4 tyres
can be tested, such a block would be incomplete. As it is not possible to test all 6 tyres on a given car.

Types of incomplete block designs:

There are two types of incomplete block designs

1. Balanced Incomplete Block Designs (BIBD)

2. Partially Balanced Incomplete Block Designs (PBIBD)

BIBD: In this type of designse every pair of treatments occur same number of times in an experiment.
Thus a BIBD is an arrangement of treatments (t) in blocks (b) each of size k (where k<t). the BIBD has
following requirements

i. Each treatment occur at most once ina a block

ii. Each treatment occur in exactly “r” blocks
iii. Each pair of treatment occur together exactly “λ” times, “λ” is an integer

Example

A BIBD for t=b=5, and r=k=4, and λ=3

(I) 1 2 3 4
(II) 1 2 3 5
(III) 1 2 4 5
(IV) 1 3 4 5
(V) 2 3 4 5
In other words we can say that if there are “t” treatments and each block can hold (k<t) treatments then
t
a BIBD can be constructed by taking ( ) blocksand assigning a different combination of treatments to
k
each block. For the above given example t=5 and k=4, so the number of blocks are ( 54)=5. The symbols
t, b, r, k, and λ are called as parameter of designs and satisfy the following relation

tr = bk…………………………………..(1)

λ (t-1) = r (k-1), ……………………(2)

r (k −1)
λ= ,
(t−1)
The above given relations represent the total number of experimental units in the design. The equation
(2) can be established by noting that a given treatment occur with (k-1) other treatments to each of “r”
blocks and also occur with each of the other (t-1) treatments in λ blocks. Also it is important to note that
for every treatments design will be incomplete. For example if we have t=b=6 and r=k=3 then
r (k −1) 3(3−1) 6
λ= , λ= = , which not an integer, so for this situation BIBD does not exist. Also it is
(t−1) (6−1) 5
not necessary if the above given relation exist (i.e. λ is an integer) it is a BIBD. For example if we have
r (k −1) 7(5−1)
t=15, b =21, r=7, and k=5 then λ= , λ= =2, is not satisfied the conditions for BIBD.
(t−1) (15−1)
Construction of Balanced Incomplete Block Design

r (k −1) 3(3−1)
Consider t=4, b=4, k=3, and r=3, then λ= , λ= =2,
(t−1) ( 4−1)
Treatments Blocks
I II III IV
1 Y Y Y N
2 N Y Y Y
3 Y N Y Y
4 Y Y N Y
So the pairs in each block is given as

I II III IV
(1,3) (1,2) (1,2) (2,3)
(1,4) (1,4) (1,3) (2,4)
(3,4) (2,4) (2,3) (3,4)
An engineer is studying the milage performance characteristics of five types of gasoline additives. In the
road test he wishes to use cars as blocks. However because of a time constraint he must use incomplete
block design with the five blocks

In these designs the allocation of treatments is laid out in the Latin Square structure. So to eliminate the
variations arisimg from two types of grouping the appropriate arrange is given below for 9-treatments in
3-units per block for 4-lattice square

Rep-I Rep-II Rep-III Rep-IV

Blk (I) (II) (III) Blk (IV) (V) (VI) Blk (VII) (VIII (IX) Blk (X) (XI) (XII)
)
(I) 1 2 3 (IV) 1 4 7 (VII) 1 86 (X) 1 9 5
(II) 4 5 6 (V) 2 5 8 (VIII 9 42 (XI) 6 2 7
)
(III) 7 8 9 (VI) 3 6 9 (IX) 5 7 3 (XII) 8 4 3
Every pair of treatment occur once in the same row and also once in the same coloumn. All comparisons
between pair of treatment have nearly equal precision.

Statistical Model

The statistical model for BIBD is given below

Y ij =μ+ α i + β j +∈ij ,

Where

μ = overall mean
α i = effect of ith treatment

β j = effect of jth block

∈ij = random error

To find the various sum of squares the procedure is as follow

2
G
CF = ,
n

Total SS=∑ ∑ Y ij −CF ,

2

i j

Total SS=Block SS +Trt SS Adj + Error SS ,

Where Trt SS Adj, is to separate the block effect and treatment effect. This adjustment is necessary
because every treatment is represent in a different set of “r” block. Difference between unadjusted
treatment total are also effected by the difference between blocks.

1
Block SS= ∑ T 2 −CF
k j .j
1
Treatment SS= ∑ T 2 −CF
r i i.
k
Trt SS Adj= ∑
λt i
2
Qi

Where

1
Q i=Y i − ∑n Y ,
k j j j
Where

1if the ith treatment appear ∈the jth block

n j={
0 Otherwise

Error SS=Total SS−Block SS−Trt SS Adj,

ANOVA table for BIBD

SOV Df SS MS F
 Trt Adj t-1
Block b-1
Error n-t-b+1
Total n-1
Note: if t = b (i.e. no of treatments is equal to the number of blocks) then the design is claaed as
symmetric which means symmetric balanced incomplete block design.

Example

Treatments Blocks Y i.
1 2 3 4 5
1 - 17 14 13 12 56
2 14 14 - 13 10 51
3 12 - 13 12 9 46
4 13 11 11 12 - 47
5 11 12 10 - 8 41
Y.j 50 54 48 50 39 241
t = 5, b = 5, k = 4, r = 4, n = bk = 5 X 4 = 20

r (k −1) 4 (4−1)
λ= = 3 , which Is an integer.
(t−1) (5−1)
2 2
G 241
CF= = =2904.05,
n 20

Total SS=∑ ∑ Y ij −CF =( 14) +(12) +(13) + …+ ( 8 ) −2904.05=76.25 ,

2 2 2 2 2

i j

1 1
Block SS= ∑
k j
2
Y . j−CF= ¿] – 2904.05 = 31.2,
4
k 4
Trt SS Adj= ∑
λt i
2
Qi =
3(5)
[ ( 8.25 )2+ ( 2.75 )2+ (−0.75 )2 + (−3.5 )2 + (−6.75 )2 ]=35.73,

Error SS=Total SS−Block SS−Trt SS Adj

Error SS=76.25−31.2−35.73=10.02
Where

1
Q i=Y i . − ∑n Y ,
k j j .j

( 14 )∗[0∗( 50) +1∗( 54) +1∗( 48 )+1∗( 50) +1∗( 39) ] = 8.25,
Q 1=56−

Q =51−( )∗[1∗( 50 )+1∗( 54 )+ 0∗( 48 ) +1∗( 50 ) +1∗( 39 )] = 2.75

1
2
4
 ( 14 )∗[1∗( 50 )+ 0∗( 54 ) +1∗( 48) +1∗(50 )+ 1∗( 39) ] = -0.75
Q 3=46−

Q =47−( )∗[1∗( 50 ) +1∗( 54 ) +1∗( 48 ) +1∗( 50 ) +0∗( 39 ) ] = -3.5

1
4
4
1
Q 5=41−( )[1∗( 50 ) +1∗ (54 ) +1∗( 48 ) +0∗( 50 ) +1∗( 39 ) ] = -6.75,
4
ANOVA table for BIBD

SOV df SS MS F Ftab
4 35.73 8.9325 9.806137725 3.35669
Block 4 31.2 7.8 8.562874251 3.35669
Error 11 10.02 0.910909
Total 19 76.25
Reject the null hypothesis.

PARTIALLY BALANCED INCOMPLETE BLOCK DESIGNS

Balanced designs may not be constructed for every experimental situation requiring incomplete blocks.
In some cases the required number of replications may not be possible. So partially balanced incomplete
block designs requiring much less replications can be constructed. The minimum number of replications
r (k −1) λ (t−1)
required for balanced designs is λ= or r = suppose we have an experiment with t = 6
(t−1) ( k−1)
treatments with k=4 which is block size means that there maximum 4 experimental units in a block
which is less than complete set of treatments (i.e. t=6), the balanced design requires 15 blocks with r =
10, and we need 60 experimental units as n = rt = 10*06 =60. It is very difficult or impossible that 60
experimental units are not available or the experiment is so much expensive. To overcome this problem
we can use partially balanced incomplete block designs. In these designs the variance of every estimated
elementary contrast among treatment effect is not the same. In these designs some pairs of treatments
occur λ 1 times, some appear λ 2 times and so on, and the remaining pairs appear λ m times.

Association scheme of the treatments

Let we have a set of treatments “t” (1, 2, …, t) a relation satisfying the following conditions is called as
m-class association sceme where m ≥ 2

1. any two symbols areeither ist, 2nd, or mth associate the relation of association being symmetric. i.e. if
the symbol α is the ith associate of β, then β is ith associate of α.

2. each symbol α has ni ith associate. The number ni being independent of α.

3.if any two symbols α and β are ith associate then the number of symbols that are jth associate of α
2
and kth associate of β is p jk and is the independent of the pair of ith associates α and β.
2
The quantities t, ni and p jk are called as the parameter of association scheme.

How the association scheme is applied in PBIBD

The association scheme of “t” treatments in “b” blocks can be arranged as follow

*every symbol occur at most are in a block

*every symbol occur exactly in “r” blocks

*if the two symbols are ith associate then they occur together in λ i blocks. The number of λ i being
independent of the particular pair of ith associates α and β.

*the number t, b, r, k, λ i, and ni are called parameters of ist kind and it can be easily seen that
t

∑ ni λ i=r (k−1)
i=1

Consider the PBIBD for six treatments in blocks of four units. As discussed above that balanced design
requires 10 replications for balance with λ=6 the PBIBD requires 12 experintal units in three blocks
with two replicates using the block size of 4 units shown in the following table

Block I 1 4 2 5
Block II 2 5 3 6
Block III 3 6 1 4
(1, 4) occur two times.

(2, 5) occur two times.

(3, 6) occur two times.

All other pairs occur once in all blocks.

Example

Consider 12 treatments (1, 2, 3, …, 12), let there are three groups based on 4 treatments each as follow

(1, 2, 3, 4) (5, 6, 7, 8) (9, 10, 11, 12)

Now we define

1) any two treatments are first associate if they belongs to the same group

2) any two tretments are 2nd associate if they belong to different groups

Now we have t = 12, n1=3, and n2 =8,

1 1 1 1 2 2 2 2
p11 =2, p12=0 , p21=0 , p22=8 , p11 =0, p12=3, p21 =3, p22=4, .
i
It can be verified that these values of p jk remains unchanged for any choice of two first and second
associate. These parametes are usually written in the form of the following matrices.
 1 2 0
p jk=[ ],
0 8
2 0 3
p jk=[ ],
3 4
Question

Consider the following PBIBD,

Blocks Treatments

1 1 2 3
2 3 4 5
3 2 5 6
4 1 2 4
5 3 4 6
6 1 5 6
Find parameters of first kind and parameters of second kind of PBIBD.

And also verify the following relationships among the parameters

1- n1 +¿ n2 =t−1,
1 2
2- p11 + p12=n 1,
1 1
3- - p11 + p12=n 1−1,
1 2
4- n1 p11=n2 p11,

5- n1 λ1 +n 2 λ 2=r (k −1),
1 1
6- p21+ p 22=n2,
2 2
7- p21+ p 22=n2−1,
1 2
8- n1 p22=n2 p 12,

Analyse the data test the, test the significance difference between the treatment means by using
adjusted treatment SS.

Solution

We have, t=6, b=6, k=3, and r =3

r (k −1) 3(3−1) 6
λ= , λ= = ,
(t−1) (6−1) 5
which is not integer so PBIBD is possible, it means that pair of treatment do not occur same number of
times. The treatment pairs are,

Blocks Treatments Treatment pairs

1 1 2 3 (1, 2) (1, 3) (2, 3)

2 3 4 5 (3, 4) (3, 5) (4,5)
3 2 5 6 (2, 5) (2, 6) (5, 6)
4 1 2 4 (1, 2) (1, 4) (2, 4)
 5 3 4 6 (3, 4) (3, 6) (4, 6)
6 1 5 6 (1, 5) (1, 6) (5, 6)
The pairs (1, 2), (3, 4), and (5, 6) occur two times it means λ 1=2.

All other pairs occur one times so λ 2=1. So ot is the PBIBD with two associates.
 Factorial Experiments
In the last semester we discuss single factor experiment

What is factor?

A quantity which influences or contribute to some variable of interest.

Such as time effect age

Fertilizer effect a crop

A specific medicine effect the health patient

Teaching method effect the marks of students

What is the effect of a factor?

What is a character? Parameter variable

What is the difference between a factor and a variable?

What is a factorial experiment?

The factorial experiment is used to investigate a number of factors (Say “P”) simultaneously or at a time,
that is we can say that several factors may affects the variable in which we are interested.

Why we need factorial experiments?

Factorials are those experiments which involve the study of effects of two or more factors.

Two factor experiment

Three factor experiment

The general p factor experiment

What is the basic requirement of factorial experiments?

The requirement for factorial experiment is that there should minimum two factors are involved in the
experiment, and each factor contains minimum two levels (Low level and High level).

What are levels?

The methods, the doses, etc of a factor is called as level.

We have three types of fertilizer (N, P, K) if we apply 50 kg/acer each then it is a single factor experiment
because the factor is fertilizer.

If we have two fertilizers (N, P) and we want to test the different rates of N and P that is no nitrogen,
and 50 kg/acer nitrogen, with no phosphorous with 50kg/acer phosphorus.

In this experiment nitrogen is an individual factor with two levels and phosphorous is another factor
with two levels.

N (0 kg/acer) Low level

N (50 kg/acer) High level

P (0 kg/acer) Low level

P (50 kg/acer) High level

If we have the following situation whether it’s a factorial experiment or not

N (50 kg/acer) High level

P (0 kg/acer) Low level

P (50 kg/acer) High level

In this situation factorial arrangements/factorial designs are not applied. This leads to a single factor
experiment.

Each factor is denoted by capital letter (A, B, C) and each level is denoted by small letter (a, b, c)

Nitrogen is a factor which is denoted by A

Phosphorous is another factor which is denoted by B

Levels are denoted by small letters as follow,

a0, and a1

for the factor A

b0 and b1

for the factor B

So when we start an experiment then we use treatment combinations which are as follow

a0 b0 (Low level of factor A and low level of factor B) this is known as control and is denoted by (I)

a1 b0 (High level of factor A and low level of factor B) this is known as effect of factor A

a0 b1 (Low level of factor A and high level of factor B) this is known as effect of factor B

a1 b1 (High level of factor A and high level of factor B) this is known as interaction effect of factor A and
factor B
These treatment combinations are known as a complete set of treatment combination in standard
order.

Example: - in comparing weight increases animals kept on different diets, animals of both genders (Male
and Female) from several breeds fed different methods may be used in factorial designs, by using each
method of feeding of each breed, it is possible to determine the best method of diet and breeds. It is
also possible to study whether the best method of feeding from diet to diet and whether the optimum
method (Best Method) and diet depend upon the gender and breed of animal. That is suppose we have
two diet methods and two breeds of animals. Diet method is factor A and Breed is factor B, there are
two methods of diet as discussed above, the first method is considered as low level and second method
is considered as high level, similarly as there are two breeds the ist breed is considered as low level and
the 2nd breed is considered as high level.

So by using a factorial design we can study the manner in which may vary with changes in other
experimental factors. (The experimental factors are effect of factor A, effect of factor B and effect of
factor AB, the individual effects are known as main effects and the combinations of different levels of
different effects are called interaction effect), the interaction of the experimental factor, the factor
design by using each combination of series of treatment and experimental conditions allows the average
effect and more important their interaction with one another to be estimated simultaneously.

By this approach consistency of each effect may test and where it is shown to be dependent on the
other treatments any inter relation may be study.

If the set of factors don’t interect all the observations may be used in making treatment comparison, if
they do interect attention may have to be restricted to particular combination. The existence of
interaction can only be verified by the use of factorial experiment and the simultaneous determination
of significant interaction greatly facilitated.

The recognization of interaction are relevant, thus allows attention to be focus on these interections

STATISTICAL MODEL: -

As we discussed earlier that there are two factors A & B, with factor A have “a” and factor B have “b”
levels (a0, a1, a2, …, an and b0, b1, b2, …, bn also we can say that a = 1, 2, 3, …, a and b = 1, 2, 3, …, b), then
the observation Y ijk is the kth observation for the ith level of factor A and jth level factor B. it can be
expressed as follow

Y ijk =μ+α i+ β j+(αβ)ij +∈ijk, I is for factor A, j is for factor B, and k is for individual replicated
observation
It means there are

i = 1, 2, …., a

j = 1, 2, …, b

and k = 1, 2, …, n

and n = r*a*b, which total no of experimental units

suppose we have two factors with two levels and each treatment combination is replicated 3 times
then,

n = 3 * 2 * 2 = 12

generally, this can be presented as follow,

Y ijk ,

Factor A has a levels, Factor B has b levels and 3 replication

Factor B
1 1 2 3 … b
y 111, y 121, y 131, … y 1 b 1,
y 112, y 113, y 122, y 123, y 132, y 1 b 2, y 1 b 3,
y 133,
2 y 211, y 212, y 213, y 221, y 222, y 223, y 231, y 2 b 1, y 2 b 2, y 2 b 3,
y 232, …
y 233 ,
3 y 311, y 312, y 313, y 321, y 322, y 323, y 331, y 3 b 1, y 3 b 2, y 3 b 3 ,
Factor A
y 332, …
y 333,
. . . . . .
. . . . . .
. . . . . .
a y a 11, y a 12, y a 13, y a 21, y a 22, y a 23, y a 31, y ab 1, y ab 2, y ab 3,
y a 32, …
y a 33,
Students construct a general table in which, factor A has 2 levels, Factor B has 2 levels and 4 replications.

Where,

μ = Overall mean
α i = effect of ith level of factor A

β j = effect of jth level of factor B

(αβ )ij = interaction for the ith level of factor A and jth level of factor B
∈ijk = Random error, residual and ∈ijk NII (0 , σ 2),
Statistical Analysis (The two-factor factorial design under completely randomized designs)

ANOVA

SOV df SS MS F-cal F-tab p-value

Factor A a-1
Factor B b-1
AB (Interaction) (a-1)*(b-1)
Error By subtraction
Total r*a*b-1

CF = (GT)2/n
t r
Total SS=∑ ∑ y ij 2−CF
j=1 i=1

∑ T j2
Treatment SS= j=1 −CF
r
t

Factor A SS =
∑ A2
j =1
−CF
r∗b
t

Factor B SS =
∑ B2
j =1
−CF
r∗a
AB SS = Treatment SS – Factor A SS – Factor B SS

Treatment SS = Factor A SS + Factor B SS + AB SS

Unbiased variance
t n
2

1 j =1
∑ xj 2
( ∑ x i)
s2= [ − i=1
]
n−1 ❑ n
Error SS = Total SS- Factor A SS – Factor B SS - AB SS

There are two varieties (V) and two levels of fertilizer (F) with 4 replications. The varieties are v 1 and v2,
the fertilizer levels are f1 and f2. The data is given below with treatment combinations.

v1f1, v2f1, v1f2, v2f2

v1f1 v2f1 v1f2 v2f2

R1 3852 2864 4788 4956
R2 2606 3794 4936 5128
R3 3144 4108 4562 4150
 R4 2864 3444 4608 4990

Variety = Factor A, a = 2

Fertilizer = Factor B, b = 2

Replication = r = 4

n = r*a*b = 4*2*2 = 16

Treatment combinations = a*b = 2*2 = 4

To find ASS and BSS construct A x B sub-table as follow

Factor B (Fertilizer) Factor A (Variety) Total

v1 Total v2 Total
R1 3852 2864
R2 2606 3794
f1 12466 14210 26676
R3 3144 4108
R4 2864 3444
R1 4788 4956
R2 4936 5128
f2 18894 19224 38118
R3 4562 4150
R4 4608 4990
Total 31360 31360 33434 33434 69794
Sum the number of replications for each treatment combination.

yabr

v1f1 v2f1 v1f2 v2f2

R1 3852 (y111) 2864 (y211) 4788 (y121) 4956 (y221)
R2 2606 (y112) 3794 (y212) 4936 (y122) 5128 (y222)
R3 3144 (y113) 4108 (y213) 4562 (y123) 4150 (y223)
R4 2864 (y114) 3444 (y214) 4608 (y124) 4990 (224)

Variety X Fertilizer sub-table: (This is a two-way table)

v1 v2 Total
f1 12466 14210 26676
f2 18894 19224 38118
Total 31360 33434 64794
262391402.
CF 4198262436 3

10974169.7
Total SS 273365572 5
 Trt SS 1083870668 270967667 8576265

262660244.
V SS 2101281956 5 268842.3

270573862.
F SS 2164590900 5 8182460

VF SS 124962.25

Error SS

SOV df SS MS F-cal F-tab p-value

Variety 1 268842.25 268842.25 1.3453857 4.74722535 0.2686
Fertilizer 1 8182460.25 8182460.25 40.948045 4.74722535 0.000
VF (Interaction) 1 124962.25 124962.25 0.6253571 4.74722535 0.4444
Error 12 2397905 199825.417
Total 15 10974169.8

Treatment sum of squares is always equal to the sum of main effect sum of square and interaction sum
of squares.
An experiment was conducted with six levels of nitrogen and four levels of varieties, construct an
ANOVA in factorial arrangements.

Rep X nitrogen X variety table (Main table)

Yield no N2
v0 v1 v2 V3 v0 v1 v2 V3
Rep1 4430 3944 3464 4126 5418 6102 4768 5192
Rep2 4478 5314 2944 4482 5166 5858 6004 4604
Rep3 3850 3660 3142 4836 6432 5586 5556 4652
n2 n3
v0 v1 v2 V3 v0 v1 v2 V3
Rep1 6076 6008 6244 4546 6462 7139 5792 2774
Rep2 6420 6127 5724 5744 7056 6982 5880 5036
Rep3 6704 6642 6014 4146 6680 6564 6370 3938
n4 n5
v0 v1 v2 V3 v0 v1 v2 V3
Rep1 7290 7682 7080 1414 8452 6228 5594 2248
Rep2 7848 6594 6662 1960 8832 7387 7122 1380
Rep3 7552 6576 6320 2766 8818 6006 5480 2014
Write down all treatment combinations in standard order

N0v0, N2v0, n2v0, n3v0, N2v0, N2v1, N2v2, N2v3, n2v0, n2v1, n2v2, n2v3, n3v0, n3v1, n3v2, n3v3, n4v0,
n4v1, n4v2, n4v3, n5v0, n5v1, n5v2, n5v3

Total no of treatment combinations = a*b = 6*4 = 24

Total no of experimental units = r*a*b = 3*24 = 72

For nitrogen X Variety sub-table:

Factor B (Variety) Factor A (Nitrogen) Total

n0 N2 n2 n3 n4 n5
R1 4430 5418 6076 6462 7290 8452
v0 R2 4478 5166 6420 7056 7848 8832
R3 3850 6432 6704 6680 7552 8818
R1 3944 6102 6008 7139 7682 6228
v1 R2 5314 5858 6127 6982 6594 7387
R3 3660 5586 6642 6564 6576 6006
R1 3464 4768 6244 5792 7080 5594
v2 R2 2944 6004 5724 5880 6662 7122
R3 3142 5556 6014 6370 6320 5480
R1 4126 5192 4546 2774 1414 2248
v3 R2 4482 4604 5744 5036 1960 1380
R3 4836 4652 4146 3938 2766 2014
Total
nitrogen X variety sub-table

Variety Nitrogen Total

n0 N2 n2 n3 n4 n5
v0
v1
v2
v3
Total
 22 (2 x 2) Factorial designs
The simplest designs which are to be considered in which two factors are of interest and each factor is
set at two levels.

The levels of factors A and B are called as low level, heigh level and also they called as absent and
present.

22 (Levels)factors 25 (Levels)factors there are five factors with two levels each. 32 (Levels)factors there are two
factors with three levels each. The general design is 2k (Levels)factors

a0 b0 (Low level of factor A and low level of factor B) this is known as control and is denoted by (I)

absent factor A and absent factor B

a1 b0 (High level of factor A and low level of factor B) this is known as effect of factor A

present factor A and absent factor B

a0 b1 (Low level of factor A and high level of factor B) this is known as effect of factor B

absent factor A and present factor B

a1 b1 (High level of factor A and high level of factor B) this is known as interaction effect of factor A and
factor B

present factor A and present factor B.

two factors are said to be interact if the effect of on factor changes as the level of the other factor
changes. This is known as interaction.
Example:

Consider an investigation into the effect of concentration of reactant and presence of catalyst on the
reaction of time of a chemical process. Let the concentration of reactants is considered as Factor A and
levels of interest 15 % and 25 % catalyst being the factor B, the high level denotes the presence of
catalyst and low level denotes the absence of catalyst. The experiment is done in three replicates. The
observations are recorded as follow.

I II III Total
A low, B 28 25 27 80 a0 b0 (1) Control)
Low
A high, B 36 32 32 100 a1 b0 a
Low
A low, B 18 19 23 60 a0 b1 b
High
A High, B 31 30 29 90 a1 b1 ab
High
Now the lower case letter (1), a, b, and ab also represent the total of n represent the n replicates at the
treatment combinations, also denote the effect of a factor by a capital letter. It means “A” represents
the effect of factor “A” and “B” represents the effect of factor “B”, and “AB” represent the interaction.
They small letters represent the treatment combinations. The average is defined as the change in
response produced by a change in the levels of that factor averaged over the levels of other factor.

a−(1) a 1 b 0−a 0 b 0
The effect of “A” at the low level of “B” is = , and the effect of “A” at the high level
2n 2n
ab−b a1 b1−a0 b1
of “B” is = , similarly, you have to find the effect of level “B” at low level of “A”, and
2n 2n
effect of level “B” at high level of “A”.

The average effect of “A” is

ab−b a−(1) 1
A= + = [ ( ab−b )+ ( a− ( 1 ) ) ]
2n 2n 2n
1
¿ [a b −a b +a b −a b ],
2n 1 1 0 1 1 0 0 0
1
¿ ¿,
2n
1
¿ [(a −a )(b +b )],
2n 1 0 1 0
1
= [( a−1)(b +1)],
2n
The average effect of “B” is

1
B= [(a −a )(b +b )],
2n 1 0 1 0
1
B= [( b−1)(a +1)],
2n
The average effect of “AB” is

1
A B= [(a −a )(b −b )],
2n 1 0 1 0
1
AB = [( b−1)(a−1)],
2n
The overall average of “M” is

1
M= [(a +a )(b +b )],
4n 1 0 1 0
 1
M= [(b+1)(a+1)],
4n
Table of “+” and “-” signs (table of contrasts, also called as table of runs) is helpful for the average
effects and treatment combinations and divisors and also it is useful for finding the sum of squares of
main effects and interactions.

Effects (1) (a 0 b 0 ¿ a (a 1 b 0 ¿ b (a 0 b 1 ¿ ab (a 1 b 1 ¿ Divisor

M + + + + 4n
A - + - + 2n
B - - + + 2n
AB + - - + 2n
“+” sign shows that the effect of factor is present and “-”shows that the effect of factor is absent.

1
M= [(b+1)(a+1)],
4n
1
M= [a b +a b + a b +a b ],
4n 0 0 1 0 0 1 1 1
1
M= [(1)+a+ b+ab ],
4n
Find effect of “A”, “B”, and “AB” from the following table:

I II III Total
A low, B Low 28 25 27 80 a0 b0 (1)
A high, B 36 32 32 100 a1 b0 a
Low
A low, B 18 19 23 60 a0 b1 b
High
A High, B 31 30 29 90 a1 b1 Ab
High
330
 Effects (1) (a 0 b 0 ¿ a (a 1 b 0 ¿ b (a 0 b 1 ¿ ab (a 1 b 1 ¿ Divisor
I or (M) + + + + 4n
A - + - + 2n
B - - + + 2n
AB + - - + 2n

1
A= [ ( ab−b ) + ( a−( 1 ) ) ] ,
2n
1
A= [ ( 90−60 )+ ( 100−80 ) ]=8.33,
2(3)
1
B=
2n
[ ( ab +b ) −( a+(1)) ]=−5.00,
1
AB =
2n
[ ( b−1 )( a−1 ) ]=1.67,

[ A]2 [50 ]2 2500

SS A = k = 2 = =208.33 ,
2 ∗n 2 ∗3 12

1
SS A = k
¿]
2 ∗n
SS B=75, SS AB=8.33

1
SS B= k
¿],
2 ∗n
1
SS AB= ¿]
k
2 ∗n
There is another method to find the sum of squares for 2k factorial designs known Yate’s method.

Total Col-I Col-II Square Diviso SS

of col-II r

a0 b0 (1) 80 180 330 108900

k
2 ∗n
2
2 ∗3=12 9075
a1 b0 a 100 150 50 2500 k
2 ∗n
2
2 ∗3=12 208.33
a0 b1 b 60 20 -30 900 k
2 ∗n
2
2 ∗3=12 75
a1 b1 ab 90 30 10 100
k
2 ∗n
2
2 ∗3=12 8.33

ANOVA table

SOV df SS MS F-cal F-tab p-value

Factor A 1 208.75
Factor B 1 75
AB (Interaction) 1 8.33
Error 8 3.92
Total 11 323
If we have three factors with two levels A (a0, a1), B (b0, b1), C (c0, c1) then the design is 23 factorial
designs. There are 8 treatment combinations.

Total Col-I Col-II Col-III Col-III

Square
a0 b0 c0 (1) 27 48 102 176 30976
a1 b0 c0 a 21 54 74 0 0
a0 b1 c0 b 25 42 -2 -4 16
a1 b1 c0 ab 29 32 2 4 16
a0 b0 c1 c 19 -6 6 -28 784
a1 b0 c1 ac 23 4 -10 4 16
a0 b1 c1 bc 17 4 10 -16 256
a1 b1 c1 abc 15 -2 -6 -16 256
 Fixed Effect Model
Statistical Model

The basic Statistical model for RCB-Designs is

Y ij =μ+ α j + β i +(αβ )ij +∈ijk

i = 1, 2 , …, a (Blocks)

j = 1, 2, …, b (Treatments)

k = 1, 2, 3, …, r

With the assumptions that different levels of Factor A and different levels of Factor B are chosen by the
experimenter. i.e. they are fixed.
2
∈ijk NII (0 , σ )

E ( ∈2ij )=σ 2, E ( ∈ij )=0, E ( ∈ij ∈gh ) =0 , wherei ≠ g ,∧ j ≠ h

In case of fixed effect models the treatment effects are fixed such that
a b b a

∑ α j=0, ∑ β i=0, ∑ (αβ )ij=0, ∑ (αβ )ij=0

i=1 j=1 j=1 i=1

Let

Y i .., Y . j ., Y ..., are the means of rows, columns, overall respectively. In such a way that,
❑ ❑

Y i ..
∑ ∑ Y ijk ,for i = 1, 2, 3, …, a
j k
Y i ..= =
br br
❑ ❑

Y . j.
∑ ∑ Y ijk , for j = 1, 2, 3, …, b
j k
Y . j .= =
ar ar
Y ...
Y ... = , for K = 1, 2, 3, …, r
abr
The total sum of square can be written as
a b r a b r
Total SS = ∑ ∑ ∑ (Y ijk −Y ... ) = ∑ ∑ ∑ (Y ijk −Y i ..+ Y i ..−Y . j . +Y . j . −Y ij. +Y ij .−Y ... + Y ...−Y ... )
2 2

i=1 j=1 k=1 i=1 j=1 k=1

a b r
= ∑ ∑ ∑ [(Y i..−Y ...)+(Y . j .−Y ... )+(Y ij .−Y i..−Y . j. +Y ... )+(Y ijk −Y ... )]
2

i=1 j=1 k=1

By applying summation and taking square, we get

 a b a b a b r
Total SS=br ∑ (Y i.. −Y ... ) + ar ∑ (Y . j . −Y ... ) + r ∑ ∑ (Y ij .−Y i.. −Y . j . +Y ... ) + ∑ ∑ ∑ (Y ijk−Y ... )2
2 2 2

i=1 j=1 i=1 j=1 i=1 j=1 k=1

,

All the cross products are zero, so

Total SS=SS A + SS B + SS AB + SS E

Where,
a b r a b r 2
(Y ... )
Total SS=∑ ∑ ∑ (Y ijk −Y ... ) =∑ ∑ ∑ Y −
2 2
, ijk
i=1 j=1 k=1 i=1 j=1 k=1 abr
a

a
(Y ) ,
∑ Y i2.. 2

SS A =br ∑ (Y i.. −Y ... )2 = i=1 − ...

i=1 br abr
b

b ∑ Y 2. j . (Y ... )2 ,
SS B=ar ∑ (Y . j .−Y ...) = 2 j=1
−
j=1 ar abr
a b
SS AB=r ∑ ∑ (Y ij . −Y i ..−Y . j . +Y ... )2=TSS subtable AB−SS A −SS B ,
i =1 j=1

a b

∑ ∑ Y 2ij. 2
(Y ... ) ,
i=1 j=1
TSS subtable AB = −
r abr

[ ][ ][ ]
a b a b

∑ ∑ Y ij. 2
(Y ... )2 ∑ Y i ..
(Y ...)2
2
∑ Y 2. j . (Y ... )2 ,
i =1 j=1 i=1 j=1
SS AB= − − − − −
r abr br abr ar abr

SS E=TSS−TSS subtable AB,

a b

a b r a b r ∑ ∑ Y 2ij . ,
SS E=∑ ∑ ∑ (Y ijk −Y ij . )2=∑ ∑ ∑ Y 2ijk − i=1 j =1

i=1 j=1 k=1 i=1 j=1 k=1 r

SOV df
A a-1
B b-1
AB (a-1)(b-1)
Error ab(r-1)
Total abr-1
For Mean Squares divide each sum of square on its respective df. And for f ratio divide each mean
square on error mean square.
Now we have 23 factorial designs, it means there are three factors with 2 levels each and there are 8
treatment combinations

Let A, B, and C be the three factors with levels (a0, a1) for factor A, (b0, b1) for factor B, (c0, c1) for factor C.
the treatment combinations are:

(1) a0 b0 c0
a a1 b0 c0
b a0 b1 c0
ab a1 b1 c0
c a0 b0 c1
ac a1 b0 c1
bc a0 b1 c1
abc a1 b1 c1
ANOVA for 23 factorial design

SOV df
A a-1
B b-1
AB (a-1)(b-1)
C c-1
AC (a-1)(c-1)
BC (b-1)(c-1)
ABC (a-1)(b-1)(c-1)
Error By subtraction
Total abcr-1

a0 a1
b0 b1 b0 b1 Total
c0 c1 c0 c1 c0 c1 c0 c1
R1 13 9 12 8 10 11 14 7
R2 14 10 13 9 11 12 15 8
Total
a0 b0 c0 a0 b0 c1 a0 b1 c0 a0 b1 c1 a1 b0 c0 a1 b0 c1 a1 b1 c0 a1 b1 c1
(1) c b bc a ac ab abc

AB SS = Total SS AB Sub table – Factor A SS – Factor B SS

AC SS = Total SS AC Sub table – Factor A SS – Factor C SS

BC SS = Total SS BC Sub table – Factor B SS – Factor C SS

Remarks:

1. The total sum of square has been partitioned into sum of square due to rows, sum of squares
due to columns, sum of squares due to interaction between the factors A & B, and sum of
square of error.
SST = SS rows + SS col + SS AB + SS E
2. There must be at least two replicates (r ≥ 2) to obtain the error sum of square.
3. The df’s for various SS’s is as follow

SOV df
A a-1
B b-1
AB (a-1)(b-1)
Error ab (r-1)
Total abr-1
Then each SS is divided by the respective df’s which gives mean squares (expected SS)
a
br ∑ α 2i
E ( MS A ) =E ( )
SS A
a−1
=σ 2 + i =1
a−1
,

b
ar ∑ β 2j
E ( MS B )=E ( b−1
SS
)=σ +
B 2 j =1
b−1
,

a b
r ∑ ∑ (αβ )ij
E ( MS AB )=E
(
SS AB
(a−1)(b−1)
=σ + i=1 j=1
2
)
(a−1)(b−1)
,

E ( MS E )=E
( SS E
ab(r −1) )
2
=σ ,

And the F-ratio is calculated as follow,

MS A
F−statistic for A= ,
MS E
MS B
F−statistic for B= ,
MS E
MS AB
F−statistic for AB= ,
MS E
 Random Effect Model
Or
Component of variance model
Statistical Model

The basic Statistical model for CRD-Designs in factorial arrangements is expressed as:

Y ijk =μ+α i+ β j+(αβ)ij +∈ijk

i = 1, 2 , …, a (Factor A)

j = 1, 2, …, b (Factor B)

k = 1, 2, 3, …, r (Replications)

in this model different levels of factor A and factor B are chosen at random from a large population. This
means that α i, β j , (αβ )ij , are distributed randomly with the following assumptions.

α i’s ~ NII (0, σ 2α ¿,

β j ’s ~ NII (0, σ 2β ¿,

(αβ )ij ’s ~ NII (0, σ 2αβ ¿,

∈ijk ’s ~ NII (0, σ 2 ¿,

And the variance of any observation is:

2 2 2 2
Var ( Y ijk ) =σ α +σ β + σ αβ + σ ,

Where these variances are called as variance components and the hypothesis for these are given as
follow:
2
H 0 :σ α,
2
H 0 :σ β,
2
H 0 :σ αβ,
The basic ANOVA remains unchanged in random effect model (i.e. same as fixed effect model) and
various sum of squares are calculated in the same way as that of fixed effect model.

The expected mean squares can be found as follow:

2 2 2
E ( MS A ) =σ +r σ αβ +br σ α ,
2 2 2
E ( MS B )=σ +r σ αβ + ar σ β ,
,
2 2
E ( MS AB )=σ +r σ αβ ,

,
2
E ( MS E )=σ ,

And the F-ratio is calculated as follow,

MS A
F−statistic for A= ,
MS E
MS B
F−statistic for B= ,
MS E
MS AB
F−statistic for AB= ,
MS E
 General 2k Factorial Design
Factorial designs are widely used in experiments involving several factors where it is necessary to study
the joint effect of these factors on a response. The general factorial designs are important as they are
widely used in research widely used in research work and also because they form the basis of other
designs of considerable practical value.

The first of these cases is based on “k” factors with each have two levels. A complete replicate of such
designs required 2x2x2x … x2 (k-times) = 2k observations. (i.e. if there are two factors then we need 2x2
=4 observation. If there are 3 factors we need 2x2x2 = 8 observations and so on, these are also called as
a set of complete treatments).

The second of these cases os 3k factorial designs which means there are “k” factors with 3 level each (i.e.
if there are two factors then we need 3x3 = 9 treatment combinations and so on).

Assumptions: -

1. The factors are fixed.

2. The design is completely randomized (based on randomization).
3. The normality assumption holds.
 23 Factorial designs
In this case we have 3 factors with 2 levels each. And it contains 8 set of treatment combinations.
Graphically it is displayed as a cube (Three dimensions).

The treatment combinations for this design is

(1), a, b, ab, c, ac, bc, abc. The df are 7 for 8 treatment combinations in ANOVA. These treatment
combinations are associated with the main effects A, B, and C, two interaction effects Ab, AC, BC, three
interaction effects ABC. Now the average effect of A

1
A= 2
[ ( a−( 1 ) ) + ( ab−b ) + ( ac−c ) + ( abc−bc ) ],
2 n
1
A= 2
[ ( a−1 )+ ( b+1 ) + ( c +1 ) ],
2 n
1
B= [ ( b−1 ) + ( a+ 1 )+ ( c+ 1 ) ],
4n
1
C= [ ( c−1 )+ ( b+1 ) + ( a+1 )],
4n
1
AB= [ ( a−1 ) + ( b−1 ) + ( c +1 ) ],
4n
1
AC= [ ( a−1 ) + ( c−1 ) + ( b+1 ) ],
4n
 1
BC= [ ( a+1 ) + ( b−1 ) + ( c−1 ) ],
4n
1
ABC= [ ( a−1 ) + ( b−1 ) + ( c−1 ) ] ,
4n
The sign table for 23 factorial designs

Treatment combinations
(1) a b ab c ac bc abc
I (M) + + + + + + + +
A - + - + - + - +
B - - + + - - + +
Effects

AB + - - + + - - +
C - - - - + + + +
AC + - + - - + - +
BC + + - - - - + +
ABC - + + - + - - +

2
( Effects)
SS of effect= 3 ,
2 n ( All treatment combinations x replications ) , axbxcxr
 Split-Plot Designs
The split-plot designs is specifically suited for two factors design that has more treatment to be
accommodated by a complete block designs. In split-plot designs all the factors are not of equal
importance. Such as in an experiment of varieties and fertilizers, the variety is less important and the
fertilizer is more important. For this purpose, we divide an experimental unit into two parts, main plot
and sub plot. The less important factor applied in main plot and the most important factor applied in the
subplot. As given in the figure below:

Let we have 5 varieties (v = 5), and 6 types fertilizer (f = 6), with three blocks. So there will be 90
experimental units. We perform our experiment as follow:

For block 1

V1 V2 V3 V4 V5
N2 N6 N2 N6 N2

N2 N5 N2 N5 N2

N3 N4 N3 N4 N3

N4 N3 N4 N3 N4

N5 N2 N5 N2 N5

N6 N2 N6 N2 N6

For block 1

V1 V2 V3 V4 V5
N2 N6 N2 N6 N2

N2 N5 N2 N5 N2

N3 N4 N3 N4 N3

N4 N3 N4 N3 N4

N5 N2 N5 N2 N5

N6 N2 N6 N2 N6

For block 3

V1 V2 V3 V4 V5
N2 N6 N2 N6 N2
 N2 N5 N2 N5 N2

N3 N4 N3 N4 N3

N4 N3 N4 N3 N4

N5 N2 N5 N2 N5

N6 N2 N6 N2 N6

ANOVA for split-plot designs

SOV df SS MS F-cal F-tab

Block/Replications r-1 = 2
Varieties (Main- v-1 = 4
plot factor)
Error1 (Main-plot (r-1)(v-1) = 8
error)
Fertilizer (Sub- f-1 = 5
plot factor)
V*F (Interaction) (v-1)(f-1) = 20
Error2 (Subplot v(r-1)(f-1) =
error) 50
(Experimental
error)
Total rvf-1 = 89
For split-plot designs with two factors we have to construct two sub-tables factor A x factor B sub-table
and blocks x main plot factor sub-table.

factor A (Varieties) x factor B (Nitrogen) sub-table

Nitrogen Varieties
V1 V2 V3 V4 V5 Total
N2

N2

N3

N4

N5

N6

Total
 Rep x main plot factor (Varieties) sub-table

Nitrogen Varieties
V1 V2 V3 V4 V5 Total
R1

R2

R3

Total

CF

Total SS

Treatment SS

Var SS

Nit SS

V*N SS

E1 SS

E2 SS
Perform Split-Plot designs for four varieties and six nitrogen levels taking nitrogen levels more important
and varieties less important.

Yield Rep1 Rep2 Rep3 Total

No V1 4430 4478 3850 12758
V2 3944 5314 3660 12918
V3 3464 2944 3142 9550
V4 4126 4482 4836 13444
N2 V1 5418 5166 6432 17016
V2 6102 5858 5586 17546
V3 4768 6004 5556 16328
V4 5192 4604 4652 14448
N2 V1 6076 6420 6704 19200
V2 6008 6127 6642 18777
V3 6244 5724 6014 17982
V4 4546 5744 4146 14436
N3 V1 6462 7056 6680 20198
V2 7139 6982 6564 20685
V3 5792 5880 6370 18042
V4 2774 5036 3938 11748
N4 V1 7290 7848 7552 22690
V2 7682 6594 6576 20852
V3 7080 6662 6320 20062
V4 1414 1960 2766 6140
N5 V1 8452 8832 8818 26102
V2 6228 7387 6006 19621
V3 5594 7122 5480 18196
V4 2248 1380 2014 5642
128473 135604 130304 394381
ANOVA for split-plot designs, v=4, N=6, Rep=3

SOV df SS MS F-cal F-tab

Block/Replications r-1 = 2 1142968.361 571484.1806 1.5140852
Varieties (Main- v-1 = 3 88258498.38
plot factor) 17651699.68 46.766259
Error1 (Main-plot (r-1)(v-1) = 6 1078230
error) 137663.7139 0.36472504
Fertilizer (Sub- + 30681791.24
plot factor) 10227263.75 27.0960233
V*F (Interaction) (v-1)(f-1) = 15 69403473.04 4626898.203 12.2584637
Error2 (Subplot v(r-1)(f-1) = 13588026.97
error) 40
(Experimental
error) 377445.1937
Total rvf-1 = 71 203074758

CF = (GT)2/n = (394381)^2/72 = 2160227405

CF
t r
Total SS=∑ ∑ y ij 2−CF
j=1 i=1 203074758
Total SS
t

∑ T j2
Treatment SS= j=1 −CF
r 188343762.7
Trt SS
t

Factor A SS =
∑ A2
j =1
−CF
r∗b 88258498.38
V SS
t

Factor B SS =
∑ B2
j =1
−CF
r∗a 30681791.24
N SS
AB SS = Treatment SS – Factor A SS – Factor B SS 69403473.04
VN SS
r

∑ Ri 2
i=1
Treatment SS= −CF 1142968.361
Rep SS t=(vxn)
Error1 SS = Rep x Var subtable Total SS – Rep SS-Var 1078230
Error1 SS SS
13588026.97
Error2 SS Subtract all sum of squares from Total SS

Unbiased variance
 t n
2
∑ xj
1 j =1
2
( ∑ x i)
s2= [ − i=1
]
n−1 ❑ n
Error SS = Total SS- Factor A SS – Factor B SS - AB SS
Nitrogen X Varieties subtable

Nitrogen
N0 N2 N2 N3 N4 N5 Total
V1 12758 17016 19200 20198 22690 26102 117964
V2 12918 17546 18777 20685 20852 19621 110399
Varieties
V3 9550 16328 17982 18042 20062 18196 100160
V4 13444 14448 14436 11748 6140 5642 65858
Total 48670 65338 70395 70673 69744 69561 394381

Rep x main plot factor (Varieties) sub-table

Varieties
V1 V2 V3 V4 Total
R1

R2

R3

Total
 Split-Split-Plot designs
It is the extension of Split-plot designs. It involves minimum three factors each with minimum two levels.
The degree of precision is associated with these three factors.

Characteristics of this design are:

1. the experimental plot is divided into three parts. Main plot which contain less important factor, sub-
plot which contain more important factor, and sub-sub-plot which contain most important factor.

2. there are three levels of precision, with the main-plot factor receiving the lowest degree of precision
and the sub-sub-plot receiving the highest degree of precision.

Layout plan:

Let we have an experiment of three factors, nitrogen with 5 levels, management practices with 3 levels
and varieties with 3 levels. So, we need 45 experimental units (the experiment is 5X3X3 = 45) for each
block.

Randomization:

We name the factors as N (N2, n2, n3, n4, n5), M (m1, m2, m3), V (v1, v2, v3)

There are three steps in the randomization and lay out plan for split-split-plot designs.

Step1: divide the experimental are into three blocks, and randomly assigned main block treatments to
the main plots randomly and independently. In this experiment the main-plot factor is nitrogen (Less
important) sub-plot factor is management practices (More important) and sub-sub-plot factor is variety
(Most important

For block 1

N2

V1M1N2 V1M2N2 V1M3N2

V2M1N2 V2M2N2 V2M3N2

V3M1N2 V3M2N2 V3M3N2

N2

V1M1N2 V1M2N2 V1M3N2

V2M1N2 V2M2N2 V2M3N2

V3M1N2 V3M2N2 V3M3N2
N3

V1M1N2 V1M2N2 V1M3N2

V2M1N2 V2M2N2 V2M3N2

V3M1N2 V3M2N2 V3M3N2

N4

V1M1N2 V1M2N2 V1M3N2

V2M1N2 V2M2N2 V2M3N2

V3M1N2 V3M2N2 V3M3N2

N5

V1M1N2 V1M2N2 V1M3N2

V2M1N2 V2M2N2 V2M3N2

V3M1N2 V3M2N2 V3M3N2

What is the difference between three factor factorial design and split-split-plot designs? Which is more
efficient and what are the advantages and disadvantages?

ANOVA for split-split-plot design

ANOVA for split-split-plot designs, N=6, M = 3, v=3, Rep=3

SOV df SS MS F-cal F-tab

Block/Replications r-1 = 2
Nitrogen (Main n-1 = 4
plot factor)
Error1 (Main-plot (r-1)(n-1) = 8
error)
Methods (Sub- m-1 = 2
plot factor)
N*M (Interaction) (n-1)(m-1) = 8
Error2 (Subplot n(r-1)(m-1) =
error) 20
Varieties (Sub- v-1 = 2
Sub-plot factor)
N*V (n-1)(v-1) = 8
M*V (m-1)(v-1) = 4
N*M*V (n-1)(m-1)(v-
1) = 16
Sub-Sub-plot nm(r-1)(v-1) =
factor (Error3) 60
(Experimental
error)
Total rnmv-1 = 134
ANOVA for split-split-plot designs, N=6, M = 3, v=3, Rep=3

Main plot = variety

Sub-plot = method

Sub-sub-plot = nitrogen

Complete the degrees of freedom and source of variation.

SOV df SS MS F-cal F-tab

Block/Replications r-1 = 2
Variety (Main plot v-1 = 2
factor)
Error1 (Main-plot (r-1)(v-1) = 4
error)
Methods (Sub- m-1 = 2
plot factor)
V*M (Interaction) (v-1)(m-1) = 4
Error2 (Subplot v(r-1)(m-1) =
error) 12
Nitrogen (Sub- n-1 = 4
Sub-plot factor)
V*N (v-1)(n-1) = 8
M*N (m-1)(n-1) = 8
V*M*N (v-1)(m-1)(n-
1) = 16
Sub-Sub-plot vm(r-1)(n-1) =
factor 72
(Experimental
error)
Total rvmn-1 = 134

CF = (GT)2/n =(930773)^2/135 6417321315

CF
t r
Total SS=∑ ∑ y ij 2−CF
j=1 i=1 2438227466
Total SS
t

∑ T j2
Treatment SS= j=1 −CF
r
Trt SS (VMN subtable total SS )
t

Variety SS =
∑V2
j=1
−CF
r∗m∗n
V SS
 t

Methods SS =
∑ M2
j=1
−CF
r∗v∗n
M SS
t

Nitrogen SS =
∑ N2
j=1
−CF
r∗v∗m
N SS
VM SS = VM subtable total SS – Variety SS – Method SS
VM SS
VN SS = VN subtable total SS – Variety SS – Nitrogen SS
VN SS
MN SS = MN subtable total SS – Method SS – Nitrogen
SS
MN SS
VMN SS = VMN subtable total SS – Variety SS – Method
SS – Nitrogen SS – VM SS – VN SS – MN SS
VMN SS
r

∑ R i2
i =1
Replication SS= −CF
Rep SS v∗m∗n
Error1 SS Error1 SS = Rep x Var subtable Total SS – Rep SS-Var SS

Error2 SS Error2 SS = Rep x Met subtable Total SS – Rep SS-Met SS

Error3SS Subtract all sum of squares from Total SS

VxM subtable

M1 M2 M3 Total
V1 66201 74820 135616 276637
V2 90825 93342 103656 287823
V3 108491 122712 135110 366313
Total 265517 290874 374382 930773

M1 M2 M3 Total
V1 V1M1 V1M2 V1M3
V2 V2M1 V2M2 V2M3
V3 V3M1 V3M2 V3M3
Total

Data file: FAHEEMAGRI

Title: Analysis over years

Function: FACTOR
Experiment Model Number 15:
One Factor Randomized Complete Block Design Combined over
Locations (or Combined over Years)
Data case no. 1 to 54.

Factorial ANOVA for the factors:

Location (Var 3: Loc) with values from 1 to 3
Replication (Var 2: Rep) with values from 1 to 3
Factor A (Var 1: Genotypes) with values from 1 to 6

Variable 4: V1

Grand Mean = 94.865 Grand Sum = 5122.720 Total Count = 54

T A B L E O F M E A N S

3 2 1 4 Total
-------------------------------------------------------
1 * * 92.076 1657.360
2 * * 92.076 1657.360
3 * * 100.444 1808.000
-------------------------------------------------------
* * 1 106.533 958.800
* * 2 84.200 757.800
* * 3 105.911 953.200
* * 4 109.111 982.000
* * 5 74.844 673.600
* * 6 88.591 797.320
-------------------------------------------------------
1 * 1 105.800 317.400
1 * 2 81.133 243.400
1 * 3 110.200 330.600
1 * 4 107.333 322.000
1 * 5 66.267 198.800
1 * 6 81.720 245.160
2 * 1 105.800 317.400
2 * 2 81.133 243.400
2 * 3 110.200 330.600
2 * 4 107.333 322.000
2 * 5 66.267 198.800
2 * 6 81.720 245.160
3 * 1 108.000 324.000
3 * 2 90.333 271.000
3 * 3 97.333 292.000
3 * 4 112.667 338.000
3 * 5 92.000 276.000
3 * 6 102.333 307.000
-------------------------------------------------------
 A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 840.460 420.230 2.8889
0.1322
-3 Error 6 872.794 145.466
4 Factor A 5 9135.404 1827.081 31.8060
0.0000
5 LA 10 1900.720 190.072 3.3088
0.0053
-7 Error 30 1723.334 57.444
--------------------------------------------------------------------------
---
Total 53 14472.711
--------------------------------------------------------------------------
---

Coefficient of Variation: 7.99%

s_ for means group 1: 2.8428 Number of Observations: 18

y

s_ for means group 4: 2.5264 Number of Observations: 9

y

s_ for means group 5: 4.3759 Number of Observations: 3

y

==========================================================================
===

Variable 5: V2

Grand Mean = 57.735 Grand Sum = 3117.690 Total Count = 54

T A B L E O F M E A N S

3 2 1 5 Total
-------------------------------------------------------
 1 * * 57.415 1033.470
2 * * 59.435 1069.830
3 * * 56.355 1014.390
-------------------------------------------------------
* * 1 55.633 500.700
* * 2 55.780 502.020
* * 3 57.427 516.840
* * 4 59.670 537.030
* * 5 57.913 521.220
* * 6 59.987 539.880
-------------------------------------------------------
1 * 1 55.313 165.940
1 * 2 55.460 166.380
1 * 3 57.107 171.320
1 * 4 59.350 178.050
1 * 5 57.593 172.780
1 * 6 59.667 179.000
2 * 1 57.333 172.000
2 * 2 57.480 172.440
2 * 3 59.127 177.380
2 * 4 61.370 184.110
2 * 5 59.613 178.840
2 * 6 61.687 185.060
3 * 1 54.253 162.760
3 * 2 54.400 163.200
3 * 3 56.047 168.140
3 * 4 58.290 174.870
3 * 5 56.533 169.600
3 * 6 58.607 175.820
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 88.142 44.071 1.6391
0.2704
-3 Error 6 161.329 26.888
4 Factor A 5 154.621 30.924 2.3206
0.0677
5 LA 10 0.000 0.000 0.0000
-7 Error 30 399.780 13.326
--------------------------------------------------------------------------
---
Total 53 803.872
--------------------------------------------------------------------------
---
 Coefficient of Variation: 6.32%

s_ for means group 1: 1.2222 Number of Observations: 18

y

s_ for means group 4: 1.2168 Number of Observations: 9

y

s_ for means group 5: 2.1076 Number of Observations: 3

y

==========================================================================
===

Variable 6: V3

Grand Mean = 30.654 Grand Sum = 1655.340 Total Count = 54

T A B L E O F M E A N S

3 2 1 6 Total
-------------------------------------------------------
1 * * 31.024 558.440
2 * * 28.914 520.460
3 * * 32.024 576.440
-------------------------------------------------------
* * 1 27.433 246.900
* * 2 32.260 290.340
* * 3 32.557 293.010
* * 4 41.937 377.430
* * 5 20.990 188.910
* * 6 28.750 258.750
-------------------------------------------------------
1 * 1 27.803 83.410
1 * 2 32.630 97.890
1 * 3 32.927 98.780
1 * 4 42.307 126.920
1 * 5 21.360 64.080
1 * 6 29.120 87.360
2 * 1 25.693 77.080
2 * 2 30.520 91.560
2 * 3 30.817 92.450
2 * 4 40.197 120.590
2 * 5 19.250 57.750
2 * 6 27.010 81.030
3 * 1 28.803 86.410
3 * 2 33.630 100.890
 3 * 3 33.927 101.780
3 * 4 43.307 129.920
3 * 5 22.360 67.080
3 * 6 30.120 90.360
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 90.745 45.373 3.3692
0.1045
-3 Error 6 80.801 13.467
4 Factor A 5 2167.999 433.600 36.0922
0.0000
5 LA 10 0.000 0.000 0.0000
-7 Error 30 360.411 12.014
--------------------------------------------------------------------------
---
Total 53 2699.955
--------------------------------------------------------------------------
---

Coefficient of Variation: 11.31%

s_ for means group 1: 0.8650 Number of Observations: 18

y

s_ for means group 4: 1.1554 Number of Observations: 9

y

s_ for means group 5: 2.0011 Number of Observations: 3

y

==========================================================================
===

Variable 7: V4

Grand Mean = 10.503 Grand Sum = 567.180 Total Count = 54

 T A B L E O F M E A N S

3 2 1 7 Total
-------------------------------------------------------
1 * * 10.033 180.600
2 * * 12.143 218.580
3 * * 9.333 168.000
-------------------------------------------------------
* * 1 9.960 89.640
* * 2 9.627 86.640
* * 3 16.960 152.640
* * 4 8.960 80.640
* * 5 8.427 75.840
* * 6 9.087 81.780
-------------------------------------------------------
1 * 1 9.490 28.470
1 * 2 9.157 27.470
1 * 3 16.490 49.470
1 * 4 8.490 25.470
1 * 5 7.957 23.870
1 * 6 8.617 25.850
2 * 1 11.600 34.800
2 * 2 11.267 33.800
2 * 3 18.600 55.800
2 * 4 10.600 31.800
2 * 5 10.067 30.200
2 * 6 10.727 32.180
3 * 1 8.790 26.370
3 * 2 8.457 25.370
3 * 3 15.790 47.370
3 * 4 7.790 23.370
3 * 5 7.257 21.770
3 * 6 7.917 23.750
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 77.029 38.515 71.1630
0.0001
-3 Error 6 3.247 0.541
4 Factor A 5 463.083 92.617 127.5682
0.0000
5 LA 10 0.000 0.000 0.0000
-7 Error 30 21.781 0.726
--------------------------------------------------------------------------
---
Total 53 565.140
--------------------------------------------------------------------------
---

Coefficient of Variation: 8.11%

s_ for means group 1: 0.1734 Number of Observations: 18

y

s_ for means group 4: 0.2840 Number of Observations: 9

y

s_ for means group 5: 0.4919 Number of Observations: 3

y

==========================================================================
===

Variable 8: V5

Grand Mean = 15.129 Grand Sum = 816.960 Total Count = 54

T A B L E O F M E A N S

3 2 1 8 Total
-------------------------------------------------------
1 * * 12.111 218.000
2 * * 19.776 355.960
3 * * 13.500 243.000
-------------------------------------------------------
* * 1 15.667 141.000
* * 2 13.311 119.800
* * 3 18.533 166.800
* * 4 16.222 146.000
* * 5 14.533 130.800
* * 6 12.507 112.560
-------------------------------------------------------
1 * 1 15.667 47.000
1 * 2 14.333 43.000
1 * 3 14.000 42.000
1 * 4 9.333 28.000
1 * 5 7.000 21.000
1 * 6 12.333 37.000
2 * 1 18.667 56.000
2 * 2 14.267 42.800
2 * 3 25.600 76.800
2 * 4 24.333 73.000
 2 * 5 22.933 68.800
2 * 6 12.853 38.560
3 * 1 12.667 38.000
3 * 2 11.333 34.000
3 * 3 16.000 48.000
3 * 4 15.000 45.000
3 * 5 13.667 41.000
3 * 6 12.333 37.000
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 600.332 300.166 73.7408
0.0001
-3 Error 6 24.423 4.071
4 Factor A 5 212.489 42.498 2.9488
0.0279
5 LA 10 430.947 43.095 2.9902
0.0098
-7 Error 30 432.358 14.412
--------------------------------------------------------------------------
---
Total 53 1700.549
--------------------------------------------------------------------------
---

Coefficient of Variation: 25.09%

s_ for means group 1: 0.4755 Number of Observations: 18

y

s_ for means group 4: 1.2654 Number of Observations: 9

y

s_ for means group 5: 2.1918 Number of Observations: 3

y

==========================================================================
===
 Variable 9: V6

Grand Mean = 3.611 Grand Sum = 195.000 Total Count = 54

T A B L E O F M E A N S

3 2 1 9 Total
-------------------------------------------------------
1 * * 3.611 65.000
2 * * 3.611 65.000
3 * * 3.611 65.000
-------------------------------------------------------
* * 1 3.333 30.000
* * 2 3.333 30.000
* * 3 4.000 36.000
* * 4 4.000 36.000
* * 5 3.000 27.000
* * 6 4.000 36.000
-------------------------------------------------------
1 * 1 3.333 10.000
1 * 2 3.333 10.000
1 * 3 4.000 12.000
1 * 4 4.000 12.000
1 * 5 3.000 9.000
1 * 6 4.000 12.000
2 * 1 3.333 10.000
2 * 2 3.333 10.000
2 * 3 4.000 12.000
2 * 4 4.000 12.000
2 * 5 3.000 9.000
2 * 6 4.000 12.000
3 * 1 3.333 10.000
3 * 2 3.333 10.000
3 * 3 4.000 12.000
3 * 4 4.000 12.000
3 * 5 3.000 9.000
3 * 6 4.000 12.000
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 0.000 0.000 0.0000
-3 Error 6 0.333 0.056
4 Factor A 5 8.833 1.767 5.4828
0.0011
5 LA 10 0.000 0.000 0.0000
 -7 Error 30 9.667 0.322
--------------------------------------------------------------------------
---
Total 53 18.833
--------------------------------------------------------------------------
---

Coefficient of Variation: 15.72%

s_ for means group 1: 0.0556 Number of Observations: 18

y

s_ for means group 4: 0.1892 Number of Observations: 9

y

s_ for means group 5: 0.3277 Number of Observations: 3

y

==========================================================================
===

Variable 10: V7

Grand Mean = 3.870 Grand Sum = 208.960 Total Count = 54

T A B L E O F M E A N S

3 2 1 10 Total
-------------------------------------------------------
1 * * 5.124 92.240
2 * * 3.267 58.810
3 * * 3.217 57.910
-------------------------------------------------------
* * 1 4.180 37.620
* * 2 3.894 35.050
* * 3 4.397 39.570
* * 4 3.440 30.960
* * 5 3.719 33.470
* * 6 3.588 32.290
-------------------------------------------------------
1 * 1 5.827 17.480
1 * 2 4.803 14.410
1 * 3 6.157 18.470
1 * 4 4.683 14.050
1 * 5 4.677 14.030
1 * 6 4.600 13.800
 2 * 1 3.347 10.040
2 * 2 3.547 10.640
2 * 3 3.567 10.700
2 * 4 2.700 8.100
2 * 5 3.147 9.440
2 * 6 3.297 9.890
3 * 1 3.367 10.100
3 * 2 3.333 10.000
3 * 3 3.467 10.400
3 * 4 2.937 8.810
3 * 5 3.333 10.000
3 * 6 2.867 8.600
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 42.536 21.268 84.5209
0.0000
-3 Error 6 1.510 0.252
4 Factor A 5 5.953 1.191 7.8632
0.0001
5 LA 10 3.515 0.352 2.3217
0.0365
-7 Error 30 4.543 0.151
--------------------------------------------------------------------------
---
Total 53 58.056
--------------------------------------------------------------------------
---

Coefficient of Variation: 10.06%

s_ for means group 1: 0.1182 Number of Observations: 18

y

s_ for means group 4: 0.1297 Number of Observations: 9

y

s_ for means group 5: 0.2247 Number of Observations: 3

y

==========================================================================
===
 Variable 11: V8

Grand Mean = 36.745 Grand Sum = 1984.230 Total Count = 54

T A B L E O F M E A N S

3 2 1 11 Total
-------------------------------------------------------
1 * * 36.367 654.610
2 * * 37.501 675.020
3 * * 36.367 654.600
-------------------------------------------------------
* * 1 36.443 327.990
* * 2 36.361 327.250
* * 3 36.752 330.770
* * 4 35.922 323.300
* * 5 37.137 334.230
* * 6 37.854 340.690
-------------------------------------------------------
1 * 1 35.500 106.500
1 * 2 38.083 114.250
1 * 3 36.917 110.750
1 * 4 33.833 101.500
1 * 5 36.520 109.560
1 * 6 37.350 112.050
2 * 1 36.247 108.740
2 * 2 35.767 107.300
2 * 3 37.807 113.420
2 * 4 38.300 114.900
2 * 5 38.257 114.770
2 * 6 38.630 115.890
3 * 1 37.583 112.750
3 * 2 35.233 105.700
3 * 3 35.533 106.600
3 * 4 35.633 106.900
3 * 5 36.633 109.900
3 * 6 37.583 112.750
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
 1 Location 2 15.436 7.718 5.4123
0.0454
-3 Error 6 8.556 1.426
4 Factor A 5 20.697 4.139 1.2646
0.3048
5 LA 10 51.651 5.165 1.5780
0.1614
-7 Error 30 98.195 3.273
--------------------------------------------------------------------------
---
Total 53 194.534
--------------------------------------------------------------------------
---

Coefficient of Variation: 4.92%

s_ for means group 1: 0.2815 Number of Observations: 18

y

s_ for means group 4: 0.6031 Number of Observations: 9

y

s_ for means group 5: 1.0445 Number of Observations: 3

y

==========================================================================
===

Variable 12: V9

Grand Mean = 195.924 Grand Sum = 10579.890 Total Count = 54

T A B L E O F M E A N S

3 2 1 12 Total
-------------------------------------------------------
1 * * 195.924 3526.630
2 * * 195.924 3526.630
3 * * 195.924 3526.630
-------------------------------------------------------
* * 1 222.200 1999.800
* * 2 188.733 1698.600
* * 3 227.200 2044.800
* * 4 178.367 1605.300
* * 5 174.367 1569.300
* * 6 184.677 1662.090
 -------------------------------------------------------
1 * 1 222.200 666.600
1 * 2 188.733 566.200
1 * 3 227.200 681.600
1 * 4 178.367 535.100
1 * 5 174.367 523.100
1 * 6 184.677 554.030
2 * 1 222.200 666.600
2 * 2 188.733 566.200
2 * 3 227.200 681.600
2 * 4 178.367 535.100
2 * 5 174.367 523.100
2 * 6 184.677 554.030
3 * 1 222.200 666.600
3 * 2 188.733 566.200
3 * 3 227.200 681.600
3 * 4 178.367 535.100
3 * 5 174.367 523.100
3 * 6 184.677 554.030
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 0.000 0.000 0.0000
-3 Error 6 2.236 0.373
4 Factor A 5 23578.230 4715.646 3268.7590
0.0000
5 LA 10 -0.000 -0.000 -0.0000
-7 Error 30 43.279 1.443
--------------------------------------------------------------------------
---
Total 53 23623.745
--------------------------------------------------------------------------
---

Coefficient of Variation: 0.61%

s_ for means group 1: 0.1439 Number of Observations: 18

y

s_ for means group 4: 0.4004 Number of Observations: 9

y

s_ for means group 5: 0.6935 Number of Observations: 3

y
==========================================================================
===

Variable 13: V10

Grand Mean = 4.264 Grand Sum = 230.250 Total Count = 54

T A B L E O F M E A N S

3 2 1 13 Total
-------------------------------------------------------
1 * * 4.264 76.750
2 * * 4.264 76.750
3 * * 4.264 76.750
-------------------------------------------------------
* * 1 4.433 39.900
* * 2 4.347 39.120
* * 3 4.410 39.690
* * 4 4.447 40.020
* * 5 3.533 31.800
* * 6 4.413 39.720
-------------------------------------------------------
1 * 1 4.433 13.300
1 * 2 4.347 13.040
1 * 3 4.410 13.230
1 * 4 4.447 13.340
1 * 5 3.533 10.600
1 * 6 4.413 13.240
2 * 1 4.433 13.300
2 * 2 4.347 13.040
2 * 3 4.410 13.230
2 * 4 4.447 13.340
2 * 5 3.533 10.600
2 * 6 4.413 13.240
3 * 1 4.433 13.300
3 * 2 4.347 13.040
3 * 3 4.410 13.230
3 * 4 4.447 13.340
3 * 5 3.533 10.600
3 * 6 4.413 13.240
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E
 K Degrees of Sum of Mean F
Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 0.000 0.000 0.0000
-3 Error 6 0.553 0.092
4 Factor A 5 5.817 1.163 9.1472
0.0000
5 LA 10 -0.000 -0.000 -0.0000
-7 Error 30 3.816 0.127
--------------------------------------------------------------------------
---
Total 53 10.186
--------------------------------------------------------------------------
---

Coefficient of Variation: 8.36%

s_ for means group 1: 0.0716 Number of Observations: 18

y

s_ for means group 4: 0.1189 Number of Observations: 9

y

s_ for means group 5: 0.2059 Number of Observations: 3

y

==========================================================================
===

Variable 14: V11

Grand Mean = 25.569 Grand Sum = 1380.720 Total Count = 54

T A B L E O F M E A N S

3 2 1 14 Total
-------------------------------------------------------
1 * * 25.569 460.240
2 * * 25.569 460.240
3 * * 25.569 460.240
-------------------------------------------------------
* * 1 23.893 215.040
* * 2 26.730 240.570
* * 3 25.863 232.770
* * 4 25.827 232.440
 * * 5 25.207 226.860
* * 6 25.893 233.040
-------------------------------------------------------
1 * 1 23.893 71.680
1 * 2 26.730 80.190
1 * 3 25.863 77.590
1 * 4 25.827 77.480
1 * 5 25.207 75.620
1 * 6 25.893 77.680
2 * 1 23.893 71.680
2 * 2 26.730 80.190
2 * 3 25.863 77.590
2 * 4 25.827 77.480
2 * 5 25.207 75.620
2 * 6 25.893 77.680
3 * 1 23.893 71.680
3 * 2 26.730 80.190
3 * 3 25.863 77.590
3 * 4 25.827 77.480
3 * 5 25.207 75.620
3 * 6 25.893 77.680
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 0.000 0.000 0.0000
-3 Error 6 21.583 3.597
4 Factor A 5 40.908 8.182 0.7361
5 LA 10 0.000 0.000 0.0000
-7 Error 30 333.420 11.114
--------------------------------------------------------------------------
---
Total 53 395.911
--------------------------------------------------------------------------
---

Coefficient of Variation: 13.04%

s_ for means group 1: 0.4470 Number of Observations: 18

y

s_ for means group 4: 1.1113 Number of Observations: 9

y

s_ for means group 5: 1.9248 Number of Observations: 3

y
==========================================================================
===

Variable 15: V12

Grand Mean = 27.226 Grand Sum = 1470.180 Total Count = 54

T A B L E O F M E A N S

3 2 1 15 Total
-------------------------------------------------------
1 * * 27.226 490.060
2 * * 27.226 490.060
3 * * 27.226 490.060
-------------------------------------------------------
* * 1 27.800 250.200
* * 2 29.133 262.200
* * 3 28.367 255.300
* * 4 25.290 227.610
* * 5 25.330 227.970
* * 6 27.433 246.900
-------------------------------------------------------
1 * 1 27.800 83.400
1 * 2 29.133 87.400
1 * 3 28.367 85.100
1 * 4 25.290 75.870
1 * 5 25.330 75.990
1 * 6 27.433 82.300
2 * 1 27.800 83.400
2 * 2 29.133 87.400
2 * 3 28.367 85.100
2 * 4 25.290 75.870
2 * 5 25.330 75.990
2 * 6 27.433 82.300
3 * 1 27.800 83.400
3 * 2 29.133 87.400
3 * 3 28.367 85.100
3 * 4 25.290 75.870
3 * 5 25.330 75.990
3 * 6 27.433 82.300
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E
 K Degrees of Sum of Mean F
Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 0.000 0.000 0.0000
-3 Error 6 16.458 2.743
4 Factor A 5 113.890 22.778 9.6092
0.0000
5 LA 10 -0.000 -0.000 -0.0000
-7 Error 30 71.113 2.370
--------------------------------------------------------------------------
---
Total 53 201.461
--------------------------------------------------------------------------
---

Coefficient of Variation: 5.66%

s_ for means group 1: 0.3904 Number of Observations: 18

y

s_ for means group 4: 0.5132 Number of Observations: 9

y

s_ for means group 5: 0.8889 Number of Observations: 3

y

==========================================================================
===

Variable 16: V13

Grand Mean = 76.604 Grand Sum = 4136.640 Total Count = 54

T A B L E O F M E A N S

3 2 1 16 Total
-------------------------------------------------------
1 * * 76.604 1378.880
2 * * 76.604 1378.880
3 * * 76.604 1378.880
-------------------------------------------------------
* * 1 78.600 707.400
* * 2 77.097 693.870
* * 3 73.997 665.970
 * * 4 80.367 723.300
* * 5 72.513 652.620
* * 6 77.053 693.480
-------------------------------------------------------
1 * 1 78.600 235.800
1 * 2 77.097 231.290
1 * 3 73.997 221.990
1 * 4 80.367 241.100
1 * 5 72.513 217.540
1 * 6 77.053 231.160
2 * 1 78.600 235.800
2 * 2 77.097 231.290
2 * 3 73.997 221.990
2 * 4 80.367 241.100
2 * 5 72.513 217.540
2 * 6 77.053 231.160
3 * 1 78.600 235.800
3 * 2 77.097 231.290
3 * 3 73.997 221.990
3 * 4 80.367 241.100
3 * 5 72.513 217.540
3 * 6 77.053 231.160
-------------------------------------------------------

A N A L Y S I S O F V A R I A N C E T A B L E

K Degrees of Sum of Mean F

Value Source Freedom Squares Square Value Prob
--------------------------------------------------------------------------
---
1 Location 2 -0.000 -0.000 -0.0000
-3 Error 6 27.696 4.616
4 Factor A 5 379.062 75.812 5.5198
0.0010
5 LA 10 0.000 0.000 0.0000
-7 Error 30 412.036 13.735
--------------------------------------------------------------------------
---
Total 53 818.794
--------------------------------------------------------------------------
---

Coefficient of Variation: 4.84%

s_ for means group 1: 0.5064 Number of Observations: 18

y

s_ for means group 4: 1.2353 Number of Observations: 9

y
s_ for means group 5: 2.1397 Number of Observations: 3
y
Let we have a set of vectors

{[ ] [ ] [ ]}
1 1 1
1 , −2 , 2
1 1 4

The linear combination is

[] [ ] [][]
1 1 1 0
c 1 1 +c 2 −2 +c 3 2 = 0
1 1 4 0

In equation form

c 1 +c 2+ c 3=0

c 1−2 c 2 +2 c3 =0

c 1 +c 2+ 4 c3 =0

x 1 v 1+ x 2 v 2 + x 3 v 3 +…+ x k v k =0

Linear Independence

Linear Dependence

Suppose four variables

X 1 =Students marks∈Calculus

X 2 =Students marks∈Statistics

X 3 =Students marks∈Computer

X 4=Students marks∈ Algebra

X 5 =Average study time for all subjects

We take the information from 20 students randomly

[ ]
x1 x2 x3 x4 x5
~
X= x 11 x12 x 13 x 14 x15 the order of this matrix is 20 x 5
x 21 x22 x 23 x 24 x25

Let X is a vector of order p = 4, present it graphically

X1 = 4

X2 = 3

X3 = 2
X4 =1
Find which vectors are linearly dependent

[] [ ] [ ]
1 1 1
x 1= 2 , x 2= 0 , x 3 − 2
1 −1 1

c 1 +c 2+ c 3=0

2 c 1+ 0 c 2−2 c3 =0

c 1−c 2 +c 3=0

When

c 1=0 , c 2=0 ,c 3=0

[] [ ] [ ]
1 2 0
x 1= 1 , x 2= 5 , x 3 1
1 −1 −1

c 1 +2 c 2+ 0 c3 =0

c 1 +5 c 2+ c3 =0

c 1−c 2−c3 =0

It has a non-trivial solution

L x= √ x 12 + x 22 + x 32+ x 42

L x= √ (−1)2+(5)2+(2)2+(−2)2

L x= √ 1+25+ 4+ 4

L x= √ 34
L x=5.83

L y= √ y1 + y2 + y3 + y 4
2 2 2 2

L x= √ (4 )2+(−3)2+(0)2 +(1)2

L x= √ 16+ 9+0+1

L x= √26
L x=5.10
Variance-Covariance Matrix

[ ( ∑ x1 )
] [ ]
2
2
1 1 ( 379 )
S11 =
n−1
∑x − 2
1
n
=
10−1
19837−
10
=608.1,

[∑ (∑ x 2
] [ ]
2
1 ) 1 ( 433 )
2
= 775.6,
2
S22= x−
2 = 25729−
n−1 n 10−1 10

[∑ (∑ x 3
] [ ]
2
1 ) 1 ( 540 )
2
= 471.8,
2
S33= x−
3 = 33406−
n−1 n 10−1 10

S12=
1
n−1 [
∑ x 1 x 2−
( ∑ x 1 )( ∑ x 2 )
n ] [
=
1
10−1
17033−
( 379 ) ( 433 )
10 ]
=69.1=S21 ,

S13=
1
n−1 [∑ x 1 x 3−
( ∑ x 1 )( ∑ x 3
n
)
] [
=
1
10−1
18262−
( 379 ) ( 540 )
10 ]
=−244.2=S 31,

S23=
1
n−1 [∑ x 2 x 3−
( ∑ x 2 )( ∑ x3
n
)
] [
=
1
n−1
23474−
( 433 ) (540 )
n ]=10.2=S32,

∑ x 1=379,
∑ x 2=433,
∑ x 3=540 ,
∑ x 21=19837 ,
∑ x 22=25729 ,
∑ x 23=33406 ,
∑ x 1 x2 =17033,
∑ x 1 x3 =18262,
∑ x 2 x3 =23474,
Matrix/Matrices

A rectangular array of numbers is called as matrix

Origin of matrix is from India and China

A=[ 22 14]
Matrix is denoted by capital letters and the numbers which are called elements are denoted by small
letters i.e. a ij

The horizontal lines in the matrix are called rows

And the vertical lines in the matrix are called columns

Number of rows and columns in the matrix is called as the order of a matrix

There are different types of matrices

Null matrix

All the elements are zero

Oij =
[00 00]
Identity matrix

[ 10 01]
I ij =

They are linearly dependent

They are coorelated

Linearly dependent

Quadratically dependent

Matrices are linear combinations

Can we say that matrices are linear combinations

What is linear combination

2x+3y = 6

4x+7y = 7

In matrix form

[ 24 37] [ xy ]=[ 67]

A ⏟X = B
What is the concept of linear combination?

What is the concept of linearly dependent?

How we find linearly dependence?

A=
[ 11 30] , Find Eigen vaues and Eigen vector

λI =λ
[ 10 01]=[ 0λ 0λ]
A−λI =
[11 03]−[ 0λ 0λ]=[1−1 λ 0
3−λ ]
⌊ A−λI ⌋=( 1−λ ) ( 3− λ )- (0)(1) = 3- λ−3 λ+ λ 2=λ 2−4 λ+3
Apply Quadratic formula for a = 1, b = -4, c = 3

−b ± √b 2−4 ac , λ=3∧ λ=¿1

λ=
2a
 Normal distribution
The pdf for normal distribution is given as

1
f ( x )= exp ⁡¿
σ √2 π
Binomial Expansion

(a+ b) =
n
( n0) a b ( n1) a
n−0 0 n−1 1
b +… +( nn) a n−n
b
n

Where n is the total no of terms which are finite

(n0) ,( n1) , … ,( nn) are the coefficients of binomial expansion

Derive the pdf for Normal distribution

Also find the mean and variance for normal distribution

Positive definite matrix

A matrix is said to be positive definite if for any vector a ≠ 0 the following form exists
~
x A x >0

Type equation here .