Analysis Of Variance

ANOVA
Difference between more than two sample
means
Whether samples are drawn from
populations having the same mean
Assumptions
Each sample is drawn from a normal
population and the sample statistics tend
to reflect the characteristics of the
population
The populations from which the samples
are drawn have identical means and
variances

1
=
2
=
3
=
4

4

Computation
One way classification
Two way classification
One way classification
Single variable of interest
Ho:
1
=
2
=
3
==
k
where k is the number
of samples drawn
1) Calculate the variance between
the samples(SSB)
Calculate mean of each sample
Calculate grand mean
Difference between grand mean and each
sample mean =deviation
Square deviations
Divide total by d.f ; v=k-1, where k is the
number of samples

2) Variance within samples(SSW)
Calculate mean of each sample
Take deviation of each observation from
respective mean
Square deviations; find total
Divide by degrees of freedom; v=n-k; n is
the total no. of observations ;k is the
number of samples
3) Calculate the F ratio
F = Variance between the samples
Variance within the samples
4) Compare to the table value
If F cal > F tab, Reject Ho
To assess the significance of possible variation in the test
scores of four candidates of a company,a battery of 5 tests
was administered on them.The results are given below.
Make an Analysis of Variance of data
Marks
A B C D
8 12 18 13
10 11 12 9
12 9 16 12
8 14 6 16
7 4 8 15
Ho :
1
=
2
=
3
=
4
.
Marks
X
1
X
2
X
3
X
4
8 12 18 13
10 11 12 9
12 9 16 12
8 14 6 16
7 4 8 15
Total 45 50 60 65
Mean 9 10 12 13
Grand Mean = 11
Variance between samples
Sample 1 Sample 2 Sample 3 Sample 4
(X
1
- 11) (X
2
- 11) (X
3
- 11) (X
4
- 11)
4 1 1 4
4 1 1 4
4 1 1 4
4 1 1 4
4 1 1 4
20 5 5 20
Sum of squares between samples
(SSB)= 50
Mean sum of squares between
samples(MSB) = 50/4-1 = 50/3 = 16.7
Variance within the samples
Sample 1 Sample 2 Sample 3 Sample 4
X
1
(X
1
-X
1
)
X
2
(X
2
-X
2
)
X
3
(X
3
-X
3
)
X
4
(X
4
-X
4
)
8 1 12 4 18 36 13 0
10 1 11 1 12 0 9 16
12 9 9 1 16 16 12 1
8 1 14 16 6 36 16 9
7 4 4 36 8 16 15 4
16 58 104 30
Total sum of squares within
samples = 208
Mean sum of squares within the samples =
208 / 20 4
= 13
ANOVA Table
Source of
Variation
Sum of
Squares
Degrees Of
Freedom
Mean
Square
Between
samples
50 3 16.7
Within
samples
208 16 13
Total 258 19
F = 16.7 = 1.28
13
Tabulated value for F with v
1
= 3 and
v
2
=16; at 5 % L.O.S = 3.24
As F cal < F tab , Accept Ho
q
To test the significance of variation in the
retail prices of a commodity in three
principal cities, Mumbai, Kolkata and
Delhi, four shops were chosen at random
in each city and the prices observed in
rupees were as follows :
Mumbai 16 8 12 14
Kolkata 14 10 10 6
Delhi 4 10 8 8
Ho : there is no significant difference in the prices
of a commodity in the three cities
Sample 1 Sample 2 Sample 3
Mumbai Kolkata Delhi
x
1
x
1
2
x
2
x
2
2
x
3
X
3
2
16 256 14 196 4 16
8 64 10 100 10 100
12 144 10 100 8 64
14 196 6 36 8 64
50 660 40 432 30 244
T = sum of all observations
= 50 + 40 + 30 = 120
Correction factor = T
2
N
= 120 / 12 = 1200
SST = Total Sum
of Squares

= x1 + x2 + x3 - CF
= (660 + 432 + 244 ) 1200
= 136


Sum of squares between the samples
=
x
1
+ x
2
+ x
3
- CF
n
1
n
2
n
3

= 1250 1200
= 50



Anova Table
Source of
Variation
Sum of
Squares
Degrees Of
Freedom
Mean
Squares
Test Statistic
Between
samples
50 r-1=2 25 F = 25 / 9.55

= 2.617

Within
samples
86 n-r=9
Total 136 11
Tabulated value of F for d.f 2,9 = 4.26 at 5%
l.o.s
Accept Ho
Q
A study investigated the perception of
corporate ethical values among individuals
specialising in different facets of
marketing(on a scale of 1-10). At 5 % l.o.s
test whether there are differences in
perceptions among the three groups.
Core marketing Marketing research Advertising
6 5 6
5 5 7
4 4 6
5 4 5
6 5 6
4 4 6
Ho: No significant difference in ethical
values among the three groups.
Marketing manager Marketing Research Advertising
x
1
x
1
x
2
x
2
x
3
x
3

6 36 5 25 6 36
5 25 5 25 7 49
4 16 4 16 6 36
5 25 4 16 5 25
6 36 5 25 6 36
4 16 4 16 6 36
30 154 27 123 36 218
T = 93
Correction factor = 480.50
SST = total sum of squares

= x
1
+ x
2
+ x
3
- CF
= 14.50
SSTR = sum of squares between the
samples
=
(x
1
) + (x
2
) + (x
3
) - CF
n
1
n
2
n
3
= 7
Source of
Variation
Sum Of
Squares
Degrees of
freedom
Mean
squares
Test Statistic
Between
samples
7 2 (no. of
samples 1)
3.5 3.5/ 0.5

= 7
Within
Samples
7.5 15 0.5
Total 14.5 17 (cr-1)
Tabulated F
2,15
= 3.68 at 5 %
Reject Ho


TWO WAY CLASSIFICATION
Source of
Variation
Sum of
Squares
Degrees of
freedom
Mean
sum of
Squares
ratio
Between
samples
SS
Columns
C-1 Ssc/c-1 MSC/
MSE
Between
rows
SS Rows R-1 SSR/ r-1 MSR/
MSE
Residual SSE (c-1)(r-1) SSE/d.f
Total SST N-1 or cr-1
Q
The following table gives the number of
subscribers added by four major telecom
players in India, in the months of August,
September, October and November. The
data is given in lakhs. Find out at 5 % level
of significance
- If the four companies differ in their
performance
- Is there a significant difference between the
months ?
Company
Months Bharti BSNL Tata
Indicom
Reliance
August 6 6 2 5
September 7 6 2 3
October 7 6 6 4
November 7 8 7 4
H1 : there is no sig diff in the performance
of the 4 companies
H2 : there is no sig diff between the months
Company
Month
Months Bharti BSNL Tata
Indicom
Reliance
Total Average
August 6 6 2 5
19 4.75
September 7 6 2 3
18 4.5
October 7 6 6 4
23 5.75
November 7 8 7 4
26 6.50
Total 27 26 17 16 86 Grand
mean =
5.375
Average 6.75 6.5 4.25 4
Correction factor = 86/ 16 = 462.25
Total sum of squares = TSS
= 6 + 6 +2 +. 4 - CF
= 51.75
Sum of Squares between months
= 19 + 18 + 23 + 26 - 462.25
4 4 4 4
= 10.25
Sum of squares between companies
= 27 + 26 + 17 + 16 - 462.25
4
= 487.5 462.25
= 25.25
Source of
Variation
Sum of
Squares
Degrees
of
freedom
Mean
Sum Of
Squares
F cal F
tabulated
Between
months
10.25 3 3.42 3.42/1.81=
1.89
3.86
Between
companies
25.25 3 8.42 4.65 3.86
Error 16.25 9 1.81
Total 51.75 15
Reject H1
Accept H2


Perform a two way ANOVA on the
data
Plots of
land
Treatment
A B C D
I 38 40 41 39
II 45 42 49 36
III 40 38 42 42
Ho : No significant difference in treatments
Ho: No significant difference in plots of land
Using coding method, subtract 40
Plot of
Land
Treatment Total
A B C D
I -2 0 +1 -1 -2
II +5 +2 +9 -4 12
III 0 -2 +2 +2 +2
Total +3 0 +12 -3 +12
Correction factor = T = 12= 12
N 12
Sum of squares between treatments:
= +3 + 0 +12 + (-3) - T = 42
3 3 3 3 N
v = 4-1 =3
.
Sum of squares between plots of land:
= (-2) + 12 +2 - T= 26
4 4 4 N
v = 3-1 = 2
Total sum of squares
= (-2) + 5+0+0+2+(-2)+1+9 +2+
(-1)+(-4)+2 - T = 132
N

Source of
Variation
Sum of
Squares
Degrees of
freedom
Mean
sum of
Squares
ratio
Between
columns
42 C-1 = 3 42/3 = 14 14/10.6
= 1.32
Between
rows
26 R-1=2 26/2=13
13/10.67
=1.218
Residual 64 6 10.67
Total 132 N-1 or cr-1
=11
.
For F
3,6,0.05
= 4.76
As Fcal < Ftab Accept Ho

For F
(2,6)0.05
= 5.14
As Fcal < Ftab Accept Ho
Q
The following are the number of kilometres /
litre which a test driver with three diff types
of cars has obtained randomly on 3 diff
occasions. Examine the hypothesis that
the diff in the average mileage is due to
chance.
Car 1 15 14.5 14.8 14.9
Car 2 13 12.5 13.6 13.8 14
Car 3 12.8 13.2 12.7 12.6 12.9 13