Correlation and Path Analysis
Uploaded by
Girma FanaCorrelation and Path Analysis
Uploaded by
Girma FanaCorrelation and Path analysis
Correlation:
• Measure the linear relationship between two
variables.
Doc ID
Correlation can be
2
Doc ID
Correlation coefficient can be estimated as
Where Cov(X,Y) = The covariance between X,Y
V(x) = the variance of X
V(Y) = the variance of Y
Some important consideration about correlation
1. Correlation coefficient is dimensional and its value did not
super 1, -1<r<1
2. r= 0 not indicate absence of correlation b/n the two variables,
only reflect the absence of linear relationship b/n these
variables
3. If X and Y are independent random variable the covariance
and correlation = 0
3
Doc ID
4. In the context of regression simple correlation coefficient
correspond to square root of coefficient of determination
simple ( )
The correlation coffecient can be tested as
t=
4
Genotypic, phenotypic and environmental
Doc ID
correlation
The Knowledge of association b/n characters
• Has importance in breeding specially if
selection one of the character presented
difficulty due to low heritability and/or has
measuring and identification problem
Phenotypic correlation:-
• The correlation measured from the measurement
of the two characters in certain number of
individuals in a population.
• Phenotypic correlation
• Genetic
• Environmental causes
5
Doc ID
Genetic correlation:
• Involve an association of heritable nature that
can be used to orient the breeding program
• The cause of Genetic correlation among variables
• Pleiotropic effect of genes
• So during selection using one variable as
indirect selection variable we should take
consideration , specially if the correlation is
negative
6
Doc ID
Environmental correlation
The environment became the cause of correlation
when the two characters are influenced with the
same difference of the environmental condition
• The positive Environmental correlation
• The two characters favored or disfavored by
the environment
• Negative environmental correlation
• One character favored by the environment
whereas the other disfavored
7
Doc ID
Estimation of genotypic/phenotypic and
Environmental correlation
The estimation can be done using the appropriate
statistical model of each design
Cov(X,Y) =
The Expected mean square for the product can be
estimated for the source of variation for respective
experimental design
• This can be done by substituting the EMQ
variance by corresponding sum square product
8
Doc ID
ANOVA table
Source of GL Mean
variation square
X Y X+Y
Block b-1
Treatment g-1 MSQTx MSQTy MSQTx+y
Residue (r-1)(g-1) MSQRx MSQRy MSQRx+y
=(
=(
9
Doc ID
a) Phenotypic correlation
b) Genotypic correlation
=
Where = (-)/r
= -)/r
= -)/r
c) Environmental correlation
=
10
Doc ID
Expected mean square for the products
Source of GL Mean square for
variation the product
(MPXY)
Block b-1
Treatment g-1
Residue (r-1)(g-1)
11
Doc ID
Path analysis
Path analysis – a standardized regression analysis
• When the variable affecting the dependent variable increase,
the correlation coefficient alone did not explain perfectly
• Path analysis: used to partitioned the correlation coefficient
into direct and indirect effect (Wright 1921, 23; Li 1975)
• The estimations are obtained using the regression equation
in the variable previously standardized
12
Doc ID
Path analysis diagram
r
r
Spike/
m2
plant P
Seed
m2 P weight
P
Grain
yield U=
Residue
13
Doc ID
Variance of the dependent variable
V(Y) = +
Cov(Y,1) = =++
Cov(Y,2) = =+
Cov(Y,3) = =
1
= 1
1
14
