ANOVA (Analysis of Variance)
Analysis of variance (ANOVA) is used as a test of means
for two or more populations. The null hypothesis, typically,
is that all means are equal.
Assumptions for ANOVA test:
ANOVA test is based on the test statistic F (or variance Ratio) for
the validity of the F-test in ANOVA, the following Assumptions are
made.
1. The observations are independent.
2. Parent populations from which observations are taken are
normal.
Types of ANOVA:
1. ANOVA - one way classification
2. ANOVA Two way classification
ANOVA - one way classification
With the null hypothesis:
H0: 1 = 2 = 3,
and the alternative:
H1: 1 2 3,
ANOVA table for one-way classified data:
The ANOVA Table
Sources of
Variation
Sum of
Squares
Degrees of
Freedom
Between Treatments
4.09
3-1=2
Error
147.72
21-3=18
Total
151.81
21-1=20
Mean Squares
F-Statistic
M.S.t = St2/ k-1 F= M.S.t / M.S.E
~F(k-1,n-k)
M.S.E = Se2/ nk
S.S.t = Sum of squares due to treatments
E.S.S = Sum of squares due to error.
T.S.S = Sum of squares due to Total
K= number of treatments.
Steps:
1. Compute: G= Grand total of all the observations= X
2. Compute correction factor (CF)= G2 / n
n= total number of observations
3. Compute RSS ( raw sum of squares)= X2
(Sum of the squares of all the observations)
4. Total Sum of squares ( TSS) = RSS- C.F
5. Compute Ti = The sum of all the observations in the ith class (i=1,2,k)
6. Between treatments ( SSt) = Ti2 - C.F
ni
7. Error Sum of Squares ( ESS) = Total Sum of squares (TSS)- Between
treatments ( SSt)
Problem 1:
To test the hypothesis that the average number of days a patient is kept in the
three local hospitals say, A,B, and C is the same, a random check on the
number of days that seven patients stayed in each hospital reveals the
following.
Hospita
lA
8
5
9
2
7
8
2
Hospita
lB
4
3
8
7
7
1
5
Hospita
lC
1
4
9
8
7
2
3
Test the hypothesis at = 0.05
Problem 2:
A trucking company wishes to test the average life of each of the four brands of tyres.
The company uses all brands on randomly selected trucks. The records showing the lives
(Thousands of miles) of tyres are given in the table.
Brand
1
20
23
18
17
Brand 2
19
15
17
20
16
Brand
3
21
19
20
17
16
Brand
4
15
17
16
18
hypothesis that the average life for each brand of
tyres is the same. Assume = 0.01
Test the
ANOVA table for two-way classified data:
St2 = Sum of squares due to treatments
Sb2 = Sum of square due to blocks
Se2 = Sum of squares due to error.
ST2 = Sum of squares due to Total
The ANOVA Table
Sources of
Variation
Sum of
Squares
Degrees of
Freedom
Mean Squares
Treatments
S.S.t= St2
k-1
MSt = St2/ k-1
Bolocks
S.S.b= Sb2
h-1
MSb = Sb2/ h-1
Error
Total
S.S.E=
Se2
T.S.S=
ST2
(h-1)(k-1)
MSE
F-Statistic
F= MSt / MSE
~F(k-1,(h-1)(k-1))
F= MSb / MSE
~F(h-1,(h-1)(k-1))
= Se2/ (h-1)(k-1)
hk-1
Steps:
1. Compute: G= Grand total of all the observations= X
2. Compute correction factor (CF)= G2 / n
n=h.k = total number of observations
k= number of rows.
h= number of columns.
3. Compute RSS ( raw sum of squares)= X2
(Sum of the squares of all the observations)
4. Total Sum of squares ( TSS) = RSS- C.F
5. Compute Ri = The sum of all the observations in the ith row
6. Between treatments ( SSt) = Ri2 - C.F
h
7. Compute Ci = The sum of all the observations in the ith coloumn
8. Between blocks ( SSb) = ch2 - C.F
k
9. Error Sum of Squares ( ESS) = Total Sum of squares (TSS)- Between treatments ( SSt) Between blocks ( SSb)
Problem:
Five doctors each test five treatments for a certain disease and
observe the number of days each patient takes to recover. The
results are (recovery time in days) given in the following table.
Treatments
Doctor
s
10
14
23
18
20
11
15
24
17
21
12
20
16
19
13
17
17
20
12
15
19
15
22
Discuss the difference between:
(a)
The doctors
=0.05
(b) the treatments. use
