Chapter 18.

: Analysis of Variance (ANOVA)

18.1 Introduction:
The analysis of variance is a powerful statistical tool for tests of significance of several
populations mean. The term Analysis of Variance was introduced by Prof. R.A. Fisher to deal
with problems in agricultural research.
The test of significance based on Z-test and t-test are only an adequate procedure for
testing the significance of one or two sample means. In some situation, three or more population
mean to be consider at a time for testing. Therefore, an alternative procedure is needed for testing
these means. For ex: five fertilizers are applied to four plots of wheat and its yield on each of the
plot is given. We may be interested in finding out whether the effect of these fertilizers on the
yields is significantly different i.e. all the fertilizers application on wheat plot gives same yield or
different yield. Answer of this problem is provided by the technique of analysis of variance.
Thus basic purpose of the analysis of variance is to test the equality of several means.
Variation is inherent in nature. The total variation in any set of numerical data is due to a
number of causes which may be classified as: (i) Assignable causes and (ii) Chance causes. The
variation due to assignable causes can be detected and measured, whereas the variation due to
chance causes is beyond the control of human hand and cannot be traced separately.
Definition of ANOVA:
The analysis of variance is the systematic algebraic procedure of decomposing (i.e.
partitioning) overall variation ( i.e. total variation) in the responses observed in an experiment
into different component of variations such as treatment variation and error variation. Each
component is attributed identifiable cause or source of variation.
18.2 Assumptions of ANOVA:
For the validity of the F-test in ANOVA the following assumptions are made.
1. The effects of different factors (treatments and environmental effects) are additive in nature.
2. The observations and experimental errors are independent

constant variation i.e. ~K(0, « h )

3. Experimental errors are distributed independently and normally with mean zero and

4. Observations of character under study follow normal distribution

18.3 One-way Classification: (One-way ANOVA)

grouped into ‘k’ classes of sizes n1, n2 , …..nk respectively ( = ∑mklW k ) as given in below table.
Suppose, n observations of random variable yij ,( i = 1, 2, …… k ; j = 1,2….ni) are

The total variation in the observation Yij can be split into the following two components:
1) The variation between the classes, commonly known as treatment variation/class variation.
2) The variation within the classes i.e., the inherent variation of the random variable within the
observations of a class.

2
 The first type of variation is due to assignable causes, which can be detected and controlled
by human endeavor and the second type of variation due to chance causes which are beyond the
control of human.
Classes/groups Total Mean
ýW =
W

W
1 y11 y12 y13 ... y1n1 T1

ýh =
h

h
2 y21 y22 y23 ... y2n2 T2

ýp =
p

p
3 y31 y32 y31 ... y3n3 T3
: : : : :.... : : :
ým =
m

m
k yk1 yk2 yk3 ... yknk Tk
Grand total (GT) Grand Mean (ý)

Test Procedure: The steps involved in carrying out the analysis are:
1) Null Hypothesis (H0): H0: µ1 = µ2 = …= µk=µ
Alternative Hypothesis (H1): all µi’s are not equal (i = 1,2,…,k)
2) Level of significance (α ): Let α = 0.05 or 0.01
3) Computation of test statistic: steps

a) Find the sum of values of all the items ( = ∑mklW k)

Various sums of squares are obtained as follows.
of the given data. Let this grand total
represented by ‘GT’.
( )‡
j
b) Then correction factor (C.F.) =
c) Find Total sum of squares (TSS): ZZ = ∑mklW ∑jk
ïlW
h
kï − (L. . )
d) Find sum of squares between the classes or between the treatments (SSTr) is
ZZ = ∑mklW jt − (L. . )
‡

Where ni (i: 1,2,…..k) is the number of observations in the ith class.

e) Find the sum of squares within the class or sum of squares due to error (SSE):
SSE = TSS - SSTr
ANOVA Table:

MST= SSTr/k-1 5Z
Sources of Variation d.f Sum of squares (S.S.) M.S.S F ratio

MSE= SSE/N-k 5Zß

Between treatments k-1 SSTr
Within treatment (Error) N-k SSE
Total N-1 TSS

=
Test Statistic: Under Ho
§»k§jò ¨ ¹}
= ∼ (o − 1, K − o)
j ¹º ¹» §¹s j¹¼ ã
§»k§jò }k¹ºkj ¹º ¹» §¹s j¹ ã

3
4) Critical value of F or Table value of F:
The table value is obtained from F-table for (k-1, N-k) df at α % & denoted it as Ftab.
5) Decision criteria:
If Fcal ≥ Ftab,⇒ Reject Ho and concluded that the class means or treatment means are
significantly different ( i.e. class means are not same).
If Fcal < Ftab, ⇒ Accept Ho and concluded that the class means or treatment means are
not significantly different (i.e. class means are not equal).
18.4 Two-way Classification: (Two-way ANOVA):
Let us consider the case when there are two factors which may affect the variate yij values
under study Ex: The yield of cow milk may be affected by rations (feeds) as well as the varieties
(breeds) of the cows. Let us now suppose that the n cows are divided into ‘h’ different groups or
classes according to their breed, each group containing ‘k’ cows and then let us consider the
effect of k treatments (rations) given at random to cows in each group on the yield of milk.
Let the suffix ‘i’ refer to the treatments (rations/feeds) and ‘j’ refer to the varieties (breed
of the cow), then the yields of milk is yij (i:1,2, …..k; j:1,2….h) of n (= R×C) cows furnish the
data for the comparison of the treatments (rations) as well as varieties. The yields may be
expressed as variate values in the following k× h two way table.
Rations Breeds Total Mean

W.
1 2 3 j h
1 y11 y12 y13 ... y1h R1
2 y21 y22 y23 ... y2h R2 h.
3 y31 y32 y31 ... y3h R3 p.
i : : : yij : : :
k yk1 yk2 yk3 ... ykh Rk m.

Grand Mean (ý)

Total C1 C2 C3 C j Ch Grand total (GT)
Mean .W .h .p .ï .º

The total variation in the observation yij can be split into the following three components:
(i) The variation between the treatments (rations)
(ii) The variation between the varieties (breeds)
(iii) The inherent variation within the observations of treatments and varieties.
The first two types of variations are due to assignable causes which can be detected and
controlled by human endeavor and the third type of variation due to chance causes which are
beyond the control of human hand.
Test procedure for two -way analysis: The steps involved in carrying out the analysis are:
1. Null hypothesis (Ho):
Ho : µ1. = µ2. = ……µk. = µ. (for comparison of treatment/ rations) i.e., there is no significant
difference between rations (treatments)

4
 H1:µ.1 = µ.2 = …µ.h = µ. (for comparison of varieties/ breed and stock) i.e. there is no
significant difference between varieties ( breeds)
2. Level of significance (α): 5% or 1%
3. Test Statistic:
1) Find the sum of values of all n (=k×h) items of the given data.
Let this grand total represented by ‘GT ’.
( )‡
Š
Then correction factor (C.F.) =
2) Find the total sum of squares (TSS) ZZ = ∑mklW ∑ºïlW h
kï − (L. . )
3) Find the sum of squares between treatments or sum of squares between rows is
ZZ = ZZ = ∑mklW − (L. . )
‡
t
º
where ‘h’ is the number of observations in each row
4) Find the sum of squares between varieties or sum of squares between columns is

ZZ\ = ZZL = ∑ºïlW − (L. . )

v‡
m
where ‘k’ is the number of observations in each column.
5) Find the sum of squares due to error by subtraction: SSE = TSS - SSR - SSC
ANOVA TABLE
Sources of Variation d.f. Sum of M.S.S F ratio
squares (S.S.)
Between Treatments k-1 SSTr MST= SST/k-1 FT=MST/ MSE
Between Varieties h-1 SSVt MSV=SSV/h-1 FV=MSV/ MSE
Within treatment and varieties (k-1)(h-1) SSE MSE= SSE/N-k
(Error)
Total n-1 TSS
4 Critical values of F table (Ftab):
(i) For comparison between treatments, obtain F-table value for [k-1, (k-1) (h-1)] df at α level
of significance and denoted it as Ftab.
(ii) For comparison between Varieties, obtain F-table value for [k-1, (k-1) (h-1)] df at α level of
significance and denoted it as Ftab.
5. Decision criteria.
(i) If FT ≥ Ftab for [k-1, (k-1) (h-1)] df at α level of significance, H0 is rejected.
(ii) If FV ≥ Ftab for [h-1, (k-1) (h-1)] df at α level of significance, H0 is rejected.