3.

6 Statistical analysis
The statistical analysis was conducted on the following aspects:
i. Analysis of variance and per se performance,
ii. Scaling tests and estimation of gene effects,
iii. Estimation of heterosis
iv. Estimation of inbreeding depression
v. Estimation of heritability
vi. Estimation of genetic advance under selection
vii. Frequency analysis
viii. Correlation analysis

3.6.1 Analysis of variance

Data collected was subjected to analysis of variance (ANOVA) to test for the significant
difference between generations in a cross for various characters with a fixed effect model.
The statistical method for randomised complete block design is:

Yij = M + Bj + Ti + Eij
Where,
M = General effect
Bj = Block effect
Ti = Treatment effect
Eij = Error component

Table 2: Analysis of variance for the experimental design

Source of d.f. S.S. M.S. Expectation of mean
variation square
Replications (r-1) Sr Mr σ2e +g σ2r
Generations (g-1) Sg Mg σ2e +r σ2g
Error (r-1) (g-1) Se Me σ2e
Where,
r = number of replications
g = number of generations

1
3.6.2 Scaling tests and estimation of gene effects
The crosses showing significant differences among entries (progenies) for the character
were subjected to generation mean analysis for estimation of gene effects using six
parameter model as suggested by Hayman (1958) and Jinks and Jones (1958).

3.6.2.1 Scaling test

The scaling tests as described by Hayman and Mather (1955) were used to check the
adequacy of the additive-dominance model for different characters in each cross. The
adequacy of scale must satisfy two conditions viz: additivity of gene effects and
independence of heritable component from non-heritable ones. The test of first condition
provides information regarding the absence or presence of gene interaction. If one of four
scaling tests was found significant, it indicated presence of epistatic and inadequacy of
additive-dominance model. The A, B, C and D tests were made using the following
equations for their scales and variances.

(a) Estimates of scales

A=2 BC 1  P1 F 1

B=2 BC 2  P2 F 1

C=4 F2  2 F1 P1 P2

D=2 F 2  BC1 BC 2

(b) Estimates of variances

VA¿ 4 V (BC 1) + V ( P1 ¿ + V ( F 1 ¿

VB¿ 4 V (BC 2) + V ( P2 ¿ + V ( F 1 ¿

VC¿ 16 V (F 2) + 4 V (F 1)+ V ( P1 ¿ + V ( P2 ¿

VD¿ 4 V (F 2) + V (BC 1 )+ V ( BC 2 ¿

Where,

2
 A, B C and D are the scales and P1, P2, F 1, F 2, BC 1 and BC 2 are generated mean of a
character, while VA, VB, VC, and VD are the corresponding variances of the scales and
V(P1), V(P2), V(F1), V(F2), V(BC1) and V(BC2) are the variance of the sample means of
the respective generation.

(c) Test of significance

The corresponding standard error and t-values calculated as follows:
Standard error t-value
S.E (A) = (VA)1/2 t (A) = A/S.E. (A)
S.E (B) = (VB)1/2 t (B) = B/S.E. (B)
S.E (C) = (VC)1/2 t (C) = C/S.E. (C)
S.E (D) = (VD)1/2 t (D) = D/S.E. (D)

The calculated the values of ‘t’ were compared with table ‘t’ value at 5 per cent and 1 per
cent level of significance. The significance of any of these scales indicates the presence
of non-allelic interaction (epistasis). When the scale is adequate, the values of A, B C and
D should be zero (entire variation is due to additive and dominance effect only) within
the limits of their respective standard errors. The significance deviation of any of these
scale from zero is taken as an indication of the presence of non-allelic interaction.

i. D provides a test for ‘i’ (additive x additive) type of interaction.

ii. C provides a test for ‘l’ (dominance x dominance) type of gene interaction.
iii. ‘j’ (additive x dominance) type of interaction has no effect on C and D but it
affects A and B. The tests A and B provide evidence of ‘t’, ‘j’ and ‘l’ type of gene
interaction.

3.6.2.2 The estimation of various gene effects

(a) In presence of non-allelic interactions

Additive-dominance model (three parameters) was found adequate to calculate various

non allelic gene effects therefore six parameter model was used.

3
Various gene effects were estimated using six parameter model as suggested by Hayman
(1958) which are given as under;

Gene effect Symbol Method of estimation

Mean [ḿ] F2

Additive [ d᷆ ] BC 1−BC 2❑

Dominance [ h᷆ ] F 1 -4 F 2 - ½ P1 +2 BC 1 +2 BC 2❑

Additive x additive [ i᷆ ] 2 BC 1 +2 BC 2❑-4 F 2

Additive x dominance [ĵ] 2 BC1- P1 -2 BC 2+ P2

Dominance x dominance [l] P1+ P2+2 F 1+4 F 2-4 BC 1-4 BC 2

Where,

P1, P2, F 1, F 2, BC 1, and BC 2 are the mean values of P1, P2, F1, F2, BC1 and BC2
generations respectively.

The variances of these estimates were obtained using the following formulae:

V(ḿ) = V( F 2)

V(d᷆ ) = V( BC 1) + V ( BC 2)

V(h᷆ ) = V( F 1) + 16V( F 2) + ¼ V( P1) + ¼ V P2)+4v( BC 1)+4V( BC 2)

V(i᷆ ) = 4V( BC1) + 4V( BC 2) +16V( F 2)

V(ĵ) = V( BC1)+ ¼ V( P1)+V( BC 2)+ ¼ V( P1)

V(l) = V( P1))+V( P2))+4V( F 1)+16V( F 2)+16V( BC1)+16V( BC 2)

Where;

4
V( P1)), V( P2)), V( F 1), V( F 2), V( BC1) and V( BC 2) are the variances of means of P1, P2,
F1, F2, BC1 and BC2 generations, respectively.

The standard error of each of the gene effect was computed as follows:

S.E. (ḿ) = V(m)½

S.E. (d᷆) = V(d)½

S.E. (h᷆ ) = V(h)½

S.E. (i᷆) = V(i)½

S.E. (ĵ) = V(j)½

S.E. (l) = V(l)½

Then, the ‘t’ values were obtained as follows

t (ḿ) = m/S.E.(m)

t (d᷆) = d/S.E.(d)

t (h᷆ ) = h/S.E.(h)

t (i᷆) = i/S.E.(i)

t (ĵ) = j/S.E(j)

t (l᷆) = l/S.E(l)

The calculated values of ‘t’ were compared with table values of ‘t’ at 5 percent and 1
percent levels of significance, respectively.

(b) In absence of non-allelic interactions

In the absence of non-allelic interactions as indicated by non-significance of scaling test,

following three parameter model suggested by Jinks and Jones (1958) was used for
estimation of genetic components mentioned as below;

5
 ḿ = ½ P1+½ P2+4 F 2-2 BC1-2 BC 2

d᷆ = ½ P1- ½ P2

3 3
h᷆ = 6 BC1+6 BC 2-8 F 2- F 1- P1 - P2
2 2

The variance of the estimates were computed using the following formulae;

V᷆ m = ¼ V( P1) + ¼ V( P2 ¿ + 16V( F 2)+4V( BC 1)+4V( BC 2)

V᷆ d = ¼ V( P1) + ¼ V( P2)

9 9
V᷆ h = 36V( BC 1)+36V( BC 2)+64V( F 2)+V( F 1)+ ( P1)+ V( P2)
4 4

The standard errors of the estimates were calculated using the following formulae;

S.E (m) = V(m)¼

S.E (d) = V(d)¼

S.E (h) = V(h)¼

The ‘t’ values of different of components were worked out using the following formulae:

t (m) = m / S.E (m)

t (d) = d / S.E (d)

t (h) = h/ S.E (h)

The test of significance of the gene effects was performed by comparing the calculated
value of ‘t’ with tabulated values of ‘t’ at 5 percent and 1 percent levels of significance.

3.6.3 Potence
Potence will be estimated by comparing F 1 and F2 means and is calculated by the
following formula.

1
F 1 F 2  [h]
2
6
 Potence will be tested by t-test. Non -significance of this test will indicate no difference
between F1 and F2 implying no potence.

3.6.4 Variance analysis, degree of dominance and number of effective factors

3.6.4.1 Variance analysis
Heritable [additive (D)] and non-heritable [dominance (H) and Environment (E)]
components of variance were estimated for each cross separately as per the formula
suggested by Mather (1949) and are given below.

V(F2) = 1/2D + 1/4H + E

V(B1) + V(B2) = ½ D + ½ H + 2 E

{V(P1) + V(P2) + V(F1)} / 3 = E

H
3.6.4.1 Degree of dominance=¿√
D
The average degree of dominance over all loci was determined by the square root of the
ratio between H and D (Mather, 1949). Here, D is the additive component of variance and
H is the dominance component of variance.

3.6.4.2 Numbers of effective factors

The numbers of effective factors will be estimated by the following formula (Mather,
1949),

1
K= ¿ ¿ ¿ ¿
4
Where, D = Least square estimate of component of genetic variation.

3.6.5 Estimation of heterosis

Heterosis expressed as percent increase or decrease of F 1 hybrid over its mid-parent
(relative heterosis) and over its better or superior parent (heterobeltiosis) were computed
as follow;

7
 F 1−MP ❑
Heterosis (h1) ¿ x 100
⃛
MP

F 1−BP❑
Heterobeltiosis (h2%) ¿ x 100
⃛
BP

Where,
F 1 = Mean performance of the F1 hybrid over three replications
⃛ = Mean value of the parents (P1 and P2) of a hybrid over three replications
MP

⃛ = Mean value of better parent over three replications

BP

The standard errors and calculated ‘t’ value for test of significance for heterosis and
heterbeltiosis were calculated as under:

Variable Standard error ‘t’ value

Relative heterosis S.E. (F1 – MP) = √(3Me/2r) F 1−MP

t(h1) =
S . E (F 1−MP)

Heterobeltiosis S.E. (F1-BP) = √(2Me/r) F 1−BP

t(h2) =
S . E (F 1−MP)

Where,

Me = Error mean square

r = Number of replications

The test of significance of the heterosis and heterobeltiosis were carried out by
comparing the calculated values of ‘t’ at 5 percent and 1 percent levels of significance.

3.6.6 Estimation of inbreeding depression

Inbreeding depression was computed by using the following formulae:

F1 F2
Inbreeding depression (%) = t (h1) = x 100
F1

8
 The standard error and ‘t’ value for test of significance for inbreeding depression were
estimated as under:

S.E. ( F 1 F 2) =√ ¿]

F1 F2
t ( F 1 F 2) =
S . E (F 1 F 2 )

Where;

F 1❑ = Mean value of the F 1 hybrid

F 2 = mean value of the F 3 generation

V( F 1) = Variance of the F 1generation

V( F 2) = variance of the F 2 generation

n1 = Number of observation in F 1 generation

n2 = Number of observation in F 3 generation

The significance of the inbreeding depression was tested by comparing the calculated ‘t’
value with the table ‘t’ value at 5 percent and 1 percent levels of significance.

3.6.7 Estimation of heritability

3.6.7.1 Heritability in broad sense
The broad sense heritability in percent was calculated by using formula suggested by
Warner (1952) as follows;

VF 2−VF 1
h2(b) (%) = x 100
VF 2

Where,

h2b = Heritability in broad sense

VF2 = Variance of F2 generation (Phenotypic variance)

9
 VF1 = Variance of F1 generation

VF2-VF1 = Genotypic variance

3.6.7.2 Heritability in narrow sense

The narrow sense heritability, as suggested by Warner (1952) was calculated as follows:

h2(n) (%) = ( ½ D/VF2) x 100

h2(n) = Heritability in narrow sense

D = additive genotypic variance

VF2 = phenotypic variance

Heritability percentage was categorized as demonstrated by Robinson et al (1949).

0-30% - low

31-60% - moderate

61% and above – high

3.6.8 Estimation of expected genetic advance under selection

The expected genetic advance at 5 percent level of selection intensity was estimated by
using the following formula

E.G.A = K.h2 (b). σp

Where;

h2 (b) = heritability in broad sense

σp = phenotypic standard deviation

K = selection differential = 2.06 at 5% selection pressure)

Expected genetic advance as percent of mean was estimated by the following formula:

10
 G. A
E.G.A (% of mean) = x 100
X

Where;

G.A = Genetic advance

X = mean of the character under study

The genetic advance as percent of mean was categorized as suggested by Johnson et al

(1995).

0-10% - low

11-20% - moderate

and above – high.

3.6.9 Correlation coefficients

Association among different characters at genotypic and phenotypic levels was worked

out as suggested by Searle8:

Where,

11
 cov.XY (g) and Cov.XY (p) denotes genotypic and phenotypic covariance between

characters X and Y, respectively. Var.X (g), Var.Y (g) and Var.X (p), Var.Y (p) denotes

variance for characters X and Y at genotypic and phenotypic levels, respectively.

3.6.10 Frequency analysis

To understand the distribution pattern of fodder yield in the various crosses, the field data
will be subjected to frequency analysis using 5% of the highest fodder yielding parents in
the various populations as class range. The nature of dispersion, modal peaks and
skewness observed will be used to describe the nature of gene segregation for fodder
yield.

12