STA1502/003/0/2021

Tutorial letter 003/0/2021

Statistical inference I
STA1502

Year module

Department of Statistics

STUDY UNIT 2: SUMMARY

university
Define tomorrow. of south africa
STUDY UNIT 2
2.1. Inference about the difference between two
population proportions
Comparing proportions from two independent populations is analogous to comparing means from
two independent populations.
The steps are:
Step 1

The hypotheses can be one of the following

1. H0 : P1 P2 D 0 against H1 : P1 P2 6D 0
2. H0 : P1 P2 D 0 against H1 : P1 P2 > 0
3. H0 : P1 P2 D 0 against H1 : P1 P2 < 0

where P1 and P2 are the population proportions.

Step 2

The sample proportions

X1
– PO1 D ; X 1 is the number of successes and n 1 is the size of sample 1.
n1
X2
– PO2 D ; X 2 is the number of successes and n 2 is the size of sample 2.
n2

Step 3

The test statistic Z for proportion is

PO1 PO2
ZDq
1 1
PN 1 PN C
n1 n2

PN is a pooled variance for proportion

X1 C X2
PN D
n1 C n2

2
 STA1502/003/0/2021

Step 4
The confidence interval for proportion

v
u
u PO1 1 PO1 PO2 1 PO2
t
PO1 PO2 Z C
2 n1 n2

The standard error for proportion equivalent to

v
u
u PO1 1 PO1 PO2 1 PO2
t
C
n1 n2
or s
1 1
PN 1 PN C
n1 n2

Step 5
The decision rule
Reject the null hypothesis H0 if the test statistic Z is greater than the critical value, otherwise
we do not reject H0 :
Reject the null hypothesis H0 if the p–value is less than (level of significance).
Reject the null hypothesis H0 if zero lies between the two confidence limits.

2.2 One–way Analysis of Variance:

Analysis of variance (ANOVA) is used when we have more than two independent samples.
ANOVA approach allows to compare multiple population or groups by taking samples from
each population (or group) to examine the effects of differences among two or more groups.
The total of variation among the group of treatment is denoted SST.
The total of variation within the group that measures the random variation is denoted SSE.
The total variation .SS Total/ D SST C SS E

– SST: Sum squared for treatment

X
k 2
SST D n j XN j X
jD1

3
 – SST: Sum squared error.

SS E D .n 1 1/ S12 C .n 2 1/ S22 C ::: C .n k 1/ sk2

XN 1 C XN 2 C :::X k
XD is grand mean
n

or
X nj
k X
xi j
iD1 iD1
XD
n

n D n 1 C n 2 C ::: C n j

The hypotheses equal to

H0 : 1 D 2 D ::: D k (if we have k treatment)

H1 : At least one population mean is different from the other population means
nj
k X
X 2
The SS Total D Xi j X :
jD1 iD1

Mean squares for treatment is

SST
M ST D
k 1

where k 1 is the degrees of freedom for treatment k is the number of treatment or group.
Mean squares for error is

SS E
MSE D
n k

Mean squares total is

SST otal
M S T otal D
n 1
The test statistic for one–way ANOVA is

M ST
FD
MSE

The critical value is F. ;k 1;n k/ :

4
 STA1502/003/0/2021

ANOVA SUMMARY TOTAL

Source of variation Degrees of freedom Sum of squares Mean of squares F

M ST
Treatment k 1 SST M ST FD
MSE

Error n k SS E MSE

Total n 1 SSTotal

2.3 Multiple comparisons

This method enable us to identify which treatment means are responsible for the differences.
The ANOVA test enable us to determine whether differences exist between two or more
population means.

There are three methods

1. Fisher’s least significant difference method (LSD)

2. The Bonferroni method
3. Tukey’s method

2.3.1 The Least Significant difference LSD is

s
1 1
LSD D t MSE C
2 ;.n k/ ni nj

where M S E is an unbiased estimator of the common variance of the populations we are testing.
We will conclude that i and j differ if

XN j XN j > L S D

where XN i XN j is the pairwise absolute differences given always a positive difference.

5
2.3.2 Bonferroni test
k .k 1/
The number of hypothesis to be tested is determined by C D where k is the number
2
of group or treatment.

The test statistic is

XN i XN j
Ti j D s
1 1
MSE C
ni nj

where n i and n j are the sample sizes of groups (treatments).

XN i and XN j are the sample means of groups (or treatments)
MSE is mean squares error calculated from the ANOVA table.

The critical value is

t and t
I.n k/ I .n k/
2c 2c
k .k 1/
where c D
2

2.4 Randomized Block (two–way) Analysis

of variance
The randomized block design identifies two factors: Treatment and blocks that both of which affect
the response.

The null and alternative are

H0 : 1 D 2 D ::: D k
H1 : The population means are not all the same.

The ANOVA table for the randomized block design

6
 STA1502/003/0/2021

Source of variation Degrees of freedom Sum of squares Mean of squares F

SST M ST
Treatment k 1 SST M ST D FD
k 1 MSE

SS B MSB
Block b 1 SS B MSB D FD
b 1 MSE

SS E
Error .k 1/ .b 1/ SS E MSE D
.k 1/ .b 1/

Total kb 1 SSTotal

k D number of treatment (or group)

b D number of block

X
k 2
SST D b XN X
jD1

X
k 2
SS B D b XN X
kD1

X
k X
k 2
SST otal D Xi j X
jD1 iD1