Agriculture Reference
In-Depth Information
(continued)
DETERMINANT OF POOLED VAR/COV
MATRIX
people depending upon their multiple charac-
teristics. The extension personal may like to
have groups of homogeneous respondents
which may fit well to specific extension strategy.
In medical sciences, classification of diseases
may provide a structural basis towards combat-
ing a group of diseases. Cluster analysis is a
multivariate statistical technique used to form
group of relatively homogeneous elements/
individuals/entities together. The main emphasis
in cluster analysis remains in exploring the pos-
sibility of formation of homogeneous groups
of individuals or objects based on multiple
characteristics. In all the clustering activities,
the main objective remains in bringing together
the individuals or the objects or the entities
within a group having minimum variance
among them and maximum variance between
the groups.
The main activities of cluster analysis encom-
pass three major steps: (1) to measure the dis-
tance between any two objects based on multiple
characteristics, (2) the procedure for formation
of clusters, that is, amalgamation technique, and
(3) to form the clusters based on the distances
calculated taking multiple characters into consid-
eration and following the appropriate amalgam-
ation technique. This will help in further
utilization of
¼
0.12996363E+10
DISCRIMINANT FUNCTION COEFFICIENTS L(I),S
0.0915 0.0325
0.1818
0.1363 0.0288
0.6955
4.1911
1.7772
0.7058
0.1388
L(I)xD(I) VALUES
0.0082
0.1601 2.8312 0.1948
0.3557 0.4062 1.9360
5.6790 0.2758 1.0746
L(I)xD(I)x100/DSQ VALUES
2.9954
3.4209 16.3054 47.8289 2.3231 9.0502
0.0695
1.3483 23.8443 1.6403
D-SQUARE
¼
0.11873580E+02
HOTELLING T-SQUARE
¼
0.50288100E+02
F VALUE FOR TESTING T-SQ
¼
2.012 WITH 10
AND 6 D.F.
CENTROID DISCRIMINANT SCORES FOR
GROUPS 3 AND 4 ARE
266.6579 AND
278.5314
12.11 Cluster Analysis
Researchers, particularly in the field of social
science, breeding, genetics, medical, and other
fields, always want to have groups of homoge-
neous elements so that there exist maximum var-
iation among the groups and minimum variation
among the individual elements of particular
group. This will facilitate further action-oriented
research with the elements of these groups. In
conventional breeding, the possibility of success
in a breeding program increases with the crossing
between two parents in two different groups hav-
ing maximum genetic distance between them.
The possibility of in breeding depression
increases as the distance between the parents
decreases. Clustering based on genetic distances
may help in identifying the parents for breeding
improvement program. In social sciences, an
extension specialist may try to adopt different
extension strategies
information generated through
cluster analysis.
12.11.1 Distance Measures
Let
there be
n
number of
individuals, all
p
measured for
number of characters. The first
task is to find out a suitable measure which can
be used to have an idea about the distances
between any two individuals based on the
p characters. These will be used to form the
distance matrix among the n individuals.
In literature several distance measures are
found; we discuss some of these below:
for different groups of
