The One-way Analysis of Variance (ANOVA)

Statistical comparison of at least 2 population or sample means can be performed using

the ANOVA technique or F-test. This method partitions the total variance of the variable of
interest into several sources or components.

Data Presentation:

P samples Observations (Xij) Sample size Sample Sample _

(ni) Total (Ti) Mean (Xi)
Sample 1 X11 X12 ………. X1n1 n1 T1 X1
Sample 2 X21 X22 ………. X2n2 n2 T2 X2
. . . .
. . . .
Sample p Xp1 Xp2 ………. Xpnp np Tp Xp
n GT GM

where: n = total number of observations ni= number of observation/treatment

GT = grand total
GM= grand mean = GT/n

Test Procedure:

I. Ho: There is no significant difference between treatment means

Ha: At least two of the means are not equal
II. Test stat: Do an F-test at α level of significance
III. DR: Reject Ho if Fc ≥ Fα(p-1,n-1), else accept Ho
IV. Computation:

ANOVA Table:
Source of Degree of Sum of Mean Squares
Variation (SV) Freedom (df) Squares SS) (MS) Fc Ftab
Between samples
(Trmt) t-1 TrSS MSTr MSTr/MSE Fα(t-1, n-t )
Within samples
(Error) n-t ESS MSE
Total n-1 TSS

where:
GT2 ∑Ti2
2
CF = ------ TSS = ∑X – CF ij Trss = ------- - CF
n ni

ESS= TSS-TrSS MSTr = TrSS/t-1 MSE= ESS/n-t

MSTr
Fc = ---------- Ftab: Fα(t-1, n-t )
MSE

V. Decision: Reject/accept Ho
VI. Conclusion:

1
Example 1:

The data below represent the number of hours of pain relief provided by four (4) different brands
of headache tablets administered to 20 subjects. These subjects were randomly divided into 4
groups and each was treated with a different brand.

Brands of Hours of relief using No. of subjects/ Total/ Mean

Tablets 5 subjects/tablets brand subject
A 6 7 5 8 3 5 29 5.8
B 8 9 6 9 7 5 39 7.8
C 3 1 4 5 4 5 17 3.4
D 7 3 5 1 2 5 18 3.6
n=20 GT = 103

Are there significant differences on the mean number of hours of relief provided by
the tablets among the different brands? Use 5% level of significance.

SOLUTION:

i) Ho: μA= μB= μC= μD (There is no significant difference in the mean number of hours of
relief provided by the headache tablets among the different brands)

Ha: At least two of the brands of headache tablets have different mean number of hours of
relief provided

ii) Test stat: Do an F-test at 5% level of significance

iii) DR: Reject Ho if Fc ≥ Fα(p-1,n-1), else accept Ho
iv) Computation:

1032
CF = ------ = 530.45
20

TSS = ∑Xij2 – CF = (62+ 72 +…+ 22) – 530.45 = 649- 530.45 = 118.55

∑Ti2 (292 +392 + 172 + 182)

Trss = ------- - CF = ----------------------------- - 530.45 = 2975/5 – 530.45
ni 5 = 595 – 530.45 = 64.55

ESS= TSS-TrSS = 118.55 – 64.55 = 54

TrSS 64.55 ESS 54

MSTr = ------- = --------- = 21.52 MSE= ------- = -------- = 3.375
t-1 4-1 n-t 20-4

MSTr 21.52
Fc = ---------- = -------- = 6.38 Ftab: Fα(t-1, n-t ) = F0.05 (3, 16) = 3.24
MSE 3.375
F0.01 (3, 16) = 5.29

ANOVA Table:
Source of Degree of Sum of Mean Squares Ftab
Variation (SV) Freedom (df) Squares SS) (MS) Fc 5% 1%
Between samples
(Trmt) 4-1 = 3 64.55 21.52 6.38* 3.24 5.29
Within samples
(Error) 20-4 = 16 54 3.375
Total 20-1 = 19 118.55

2
 V. Decision: Reject Ho since Fc > Ftab (6.38 > 3.24)

VI. Conclusion: At least two of the brands of headache tablets have different mean
number of hours of relief provided OR
There are significant differences in the number of hours of relief provided
among the different headache tablets

Recommendation: ?

Example 2.

Researchers who are interested in stress as it relates to work assignments in industrial

settings, conducted a study in which pulse rates of three groups of workers are compared. The
first group consists of office staff who perform routine clerical duties, the second group works
with hazardous materials while the third performs tasks similar to those performed by the second
group but do not come in contact with hazardous materials. Pulse rates are taken at the midpoint
of the one hour lunch period for all three groups. The pulse rates thus obtained are provided
below:

Office workers : 58 64 71 66 79 74 70

Hazardous materials : 88 59 74 80 81 69 90

Non-hazardous materials : 65 70 79 66 74 79 60

Test the hypothesis that the average pulse rates differed significantly among the three
groups of workers at 1% level of significance.

CHI-SQUARE TEST (χ2-test)

Tests for independence or significant relationship between two nominal or ordinal

variables. It also tests for equality of any number of proportions or frequencies in rxc
contingency table.

Data Presentation: r x c contingency table

Column 1 Column 2 . Column c

Row 1 f11 f12 f1c
Row 2 f21 f22 f2c
.
Row r fr1 fr2 frc

where: f = frequency count

Test Procedure:

i) Ho: All population proportions are equal or the same

There is no significant relationship between r and c variables
r and c variables are not significantly related/dependent
r and c variables are independent from each other

Ha: All population proportions are not equal or different

There is a significant relationship between r and c variables
r and c variables are significantly related/dependent
r and c variables are dependent from each other

ii) Test stat: Do a χ2-test at α level of significance

3
iii) DR: Reject Ho if χ2c ≥ χ2α (r-1)(c-1) else accept Ho
iv) Computation:

(Obs-Exp)2
2
χ c = ∑ --------------- vs. χ2α (r-1)(c-1)
Exp
where:

Obs = Observed values

Row total x Column total
Exp = Expected values = -------------------------------
Grand total

v) Decision: Reject/accept Ho
vi) Conclusion:

Example 1:

A study was conducted regarding nutritional instruction and low birth weight babies. In
this study, 314 mothers received nutritional instruction of which 23 had low birth weight babies.
By contrast, 39 of the 316 mothers in the non-instruction group had low birth weight babies.
With the data shown below, test the hypothesis that the proportion of low birth babies born to the
mothers receiving nutritional instruction is the same as the proportion born to mothers who did
not receive instruction.

With Instruction No Instruction

Low birth weight 23 39
Not low birth weight 291 277

SOLUTION:

i) Ho: The proportion of low birth babies born to the mothers receiving nutritional instruction
is the same as the proportion born to mothers who did not receive instruction.

Ha: The proportion of low birth babies born to the mothers receiving nutritional instruction
is not the same as the proportion born to mothers who did not receive instruction.

ii) Test stat: Do a χ2-test at α level of significance

iii) DR: Reject Ho if χ2c ≥ χ2α (r-1)(c-1) else accept Ho
iv) Computation:

Nutritional Instruction
With Instruction No Instruction Total
Low birth weight 23 (30.90) 39 (31.10) 62
Not low birth weight 291 (283.10) 277 (284.90) 568
Total 314 316 630

(314)(62) (316)(62)
Exp11 = ----------- = 30.90 Exp12 = ----------- = 31.10
630 630

(314)(568) (316)(568)
Exp11 = ------------- = 283.10 Exp12 =------------- = 284.90
630 630

4
 (Obs-Exp)2 (23-30.90)2 (39-31.10)2 (291-283.10)2 (277-284.90)2
2
χ c = ∑ --------------- = ------------- + ------------- + ----------------- + ---------------
Exp 30.90 31.10 283.10 284.90

= 2.02 + 2.01 + 0.22 + 0.22 = 4.47

χ2 tab = χ2α (r-1)(c-1) = χ20.05 (2-1)(2-1) = χ20.05 (1) = 3.841

v) Decision: Reject Ho since 4.47>3.841

vi) Conclusion: The proportion of low birth babies born to the mothers receiving nutritional
instruction is not the same as the proportion born to mothers who did not
receive instruction.

Example 2.

The side effects of nonsteroidal anti-inflammatory drugs (NSAID) include problems

involving peptic ulceration, renal function, and liver disease. In 1996, the American College of
Rheumatology issued and disseminated guidelines recommending baseline tests (CBC, hepatic
panel, and renal tests) when prescribing NSAID.A study was conducted by Rothenberg and
Holcomb (A-6)to determine if physicians taking part in a national database of computerize
medical records performed the recommended baseline tests when prescribing NSAID. The
researches classified physicians in the study into four categories; those practicing in internal
medicine, family practice, academic family practice, and multispecialty groups. The data appear
in the following table.

--------------------------------------------------------------------------------------
Performed Baseline Tests
Practice Type Yes No
--------------------------------------------------------------------------------------
Internal medicine 294 921
Family practice 98 2862
Academic family practice 50 3064
Multispecialty groups 203 2652
--------------------------------------------------------------------------------------

Do the data above provide sufficient evidence for us to conclude that type of practice and
performance of baseline tests are related? Use α=0.01.

5
 PROBLEM SET

1. A study by Sma`rason et al. was motivated by the observations of other researchers

that sera from preeclamptic women damaged cultured human endothelial cells. Subjects for
the present study were women with preeclampsia, matched control women with normal
pregnancies, and nonpregnant women of childbearing age. Among the data collected were the
following observations on a relevant variable measured on subjects in the three groups.

-----------------------------------------------------------------------------------------------
Preeclampsia Pregnant Controls Nonpregnant Controls
-----------------------------------------------------------------------------------------------
113.5 91.4 94.5
106.6 95.6 115.9
39.1 113.1 107.2
95.5 100.8 103.2
43.5 88.2 104.7
49.2 92.2 94.9
99.5 78.6 93.0
102.9 96.9 100.4
----------------------------------------------------------------------------------------------

2. Hughes et al. point out that systematic complications in acute pancreatitis are largely
responsible for mortality associated with the disease. They note further that proinflammatory
cytokines, particularly TNFα, may play a central role in acute pancreatitis by mediating the
systemic squelae. Among the data collected were the following hematocrit (%) values for
adult male rats surviving more than 48 hours:

Sham : 38 40 32 36 40 40 38 40 38 40
Untreated : 56 60 50 50 50
Treated : 40 42 38 46 36 35 40 40 55 35 36

3. Boles and Johnson examined the beliefs held by adolescents regarding smoking and weight.
Respondents characterized their weight into three categories: underweight, overweight, or
appropriate. Smoking status was categorized according to the answer to the question, ”Do
you currently smoke, meaning one or more cigarettes per day? “ The following table shows
the results of a telephone study of adolescents in the age group 12-17.

=================================
Smoking
Yes No
-------------------------------------------------------
Underweight 17 97
Overweight 25 142
Appropriate 96 816
=================================

Do the data provide sufficient evidence to suggest that weight perception and smoking status
are related in adolescents? Use α=0.05.

6
 4. A sample of 500 college students participated in a study designed to evaluate the level of
college students’ knowledge of a certain group of common diseases. The following table
shows the students classified by major field of study and level of knowledge of the group of
diseases.

==========================================
Knowledge of Diseases
Major Good Poor Total
-----------------------------------------------------------------------
Premedical 31 91 122
Other 19 359 378
----------------------------------------------------------------------
Total 50 450 500
==========================================

Do these data suggest that there is a relationship between knowledge of the group of diseases
and major field of study of the college students from which the present sample was drawn?
Let α=0.01.

5. The following table shows the results of a survey in which the subjects were a sample of
300 adults residing in a certain metropolitan area. Each subject was asked to indicate which
of three policies they favored with respect to smoking in public places.

-------------------------------------------------------------------------------------------------------------------
Policy Favored
----------------------------------------------------------------------------------
Highest No restrictions Smoking allowed No smoking No
Education Level On Smoking in designated at all opinion
areas only
------------------------------------------------------------------------------------------------------------------
College students 5 44 23 3
High school graduate 15 100 30 5
Grade school graduate 15 40 10 10
-----------------------------------------------------------------------------------------------------------------

Can one conclude from these data that, in the sampled population, there is a relationship between
level of education and attitude toward smoking in public places? Let α=2.5%

6. The purpose of a study by brown and persley (A-8) was to characterize acute hepatitis A in
patients more than 40 years old. They performed a retrospective chart review of 20 subjects
who were diagnosed with acute hepatitis A, but were not hospitalized. Of interest was the use
of age (years) to predict bilirubin levels (mg/dL). The following data were collected.

--------------------------------------------------------------------------------------------------------------
Age(Years) Bilirubin (mg/dL) Age (Years) Biliribin (mg/dL)
---------------------------------------------------------------------------------------------------------------
78 7.5 44 7.0
72 12.9 42 1.8
81 14.3 45 .8
59 8.0 78 3.8
64 14.1 47 3.5
48 10.9 50 5.1
46 12.3 57 16.5
42 1.0 52 3.5
58 5.2 58 5.6
52 5.1 45 1.9

7
7. A simple random sample of 15 apparently healthy children between the age of 6 months and
15 years yielded the following data on age, X, and liver volume p0er unit of body weight
(ml/kg), Y:

------------------------------------------------------------
X Y X Y
------------------------------------------------------------
0.5 41 10.0 26
0.7 55 10.1 35
2.5 41 10.9 25
4.1 39 11.5 31
5.9 50 12.1 31
6.1 32 14.1 29
7.0 41 15.0 23
8.2 42
---------------------------------------------------------

8. A study by Triller et al. (A-14) examine the length of time required for home health-care
nurses to repackage a patient’s medication into various medication organizers (i.e. pill
boxes). For the 19 patients in the study, researchers recorded the time required for
repackaging of medications. They also recorded the number of problems encountered in the
repackaging session.

Patient No. 1 2 3 4 5 6 7 8 9 10

No.of Problems 9 2 0 6 5 3 3 1 2 0

Repackaging
Time(Minutes) 38 25 5 18 15 25 10 5 10 15

9. The following are the weights (kg) and blood glucose levels (mg/100 ml) of 16 apparently
healthy adult males:

Weight (X) Glucose(Y)

64.0 108
75.3 109
73.0 104
82.1 102
76.2 105
95.7 121
59.4 79
93.4 107
82.1 101
78.9 85
76 7 99
82.1 100
83.9 108
73.0 104
64.4 102
77.6 87

Find the simple linear regression equation and test Ho: β=0 using both ANOVA and the t test.
Test Ho: ρ=0 and construct a 95 percent confidence interval for ρ. What is the predicted glucose
level for a man who weights 95 kg? Construct the 95 percent prediction interval for his weight. Let α =
0.5 for all tests.

8
Patient No. 11 12 13 14 15 16 17 18 19

No.of Problems 1 2 1 0 0 10 0 1 1

Repackaging
Time(Minutes) 10 15 17 18 23 29 5 22 20