J. Agric. Environ. Sci. Vol. 6 No. 1 (2021) ISSN: 2616-3721 (Online); 2616-3713 (Print)
Publication of College of Agriculture and Environmental Sciences, Bahir Dar University 19
Application of Multivariate Analysis for the Differentiation of Indigenous Goat
Populations of South Gondar, Ethiopia
Birara Tade
1
, Aberra Melesse
1*
, Simret Betsha
1
1
School of Animal and Range Sciences, College of Agriculture, Hawassa University, P. O. Box 5, Hawassa,
Ethiopia
*Corresponding author: a_melesse@uni-hohenheim.de; a_melesse@yahoo.com
Received: November 21, 2020 Accepted: February 24, 2021
Abstract: The study was carried out to describe the indigenous goat population structure in selected districts
(Fogera, Farta and Libokemkem) of South Gondar zone by applying multivariate analysis on morphometric
variables. Fourteen morphometric traits were taken from 153 male and 357 female goats. The results indicated
that the district had a significant effect on all traits of male goats except for body length (BL), height at wither
(HW), height at rump (HR), ear length (EL) and scrotal circumference. The district effect in females was also
significant for BL, heart girth (HG) and chest depth, paunch girth (PG), HR, and teat length. Age had a highly
significant effect on all traits except for EL showing a high heterogeneity among males and females of different
flocks. The cluster analysis showed two distinct groups in which Farta goats were included in one cluster while
group two included the Fogera and Libokemkem goats under one sub-cluster. The canonical discriminant
analysis indicated that Fogera and Libokemkem goats were the closest while the Farta and Fogera goats were
the furthest. However, the Mahalanobis distances between the three goat populations were too small indicating
the existence of homogeneity among them. The discriminant analysis correctly assigned the respective 58.6%,
62.3% and 63.2% of the Farta, Fogera and Libokemkem goat populations into their source population with an
overall 61.4% accuracy rate. In conclusion, multivariate analysis identified BL, HG, HW, PG, HR, canon
circumference, rump length, and width as the most imperative traits to effectively differentiate the indigenous
goat populations in the studied districts.
Keywords: Indigenous goats, Morphometric traits, Multivariate analysis, South Gondar
This work is licensed under a Creative Commons Attribution 4.0 International License
1. Introduction
Goat production in many developing nations like
Ethiopia is one of the major means of improving
the livelihoods of poor livestock keepers, reducing
poverty, and attaining sustainable agriculture and
universal food security. They are the most
important livestock species for many smallholder
farmers and pastoral communities due to their low
maintenance requirements, excellent prolificacy,
short generation interval, and potential to adapt a
wide range of agro-ecological zones (Peacock,
2005; Okpeku et al., 2011). They can easily
produce and reproduce on shrubs and trees in
adverse harsh environments (including tsetse
infected areas) where no other crops can be
cultivated. However, more than half of the local
breeds in the world are threatened and have not
been fully characterized (Arandas et al., 2017).
The population of goats in Ethiopia is estimated at
38.96 million (CSA, 2018/19). The size of goat
populations in Ethiopia has increased more rapidly
(134%) than the sheep (65%) and cattle (38%)
indicating their growing importance in the
livestock agriculture of the country. They are
commonly reared in crop-livestock and agro-
pastoral farming systems, and are widely
distributed across different agro-ecological zones
of Ethiopia (Gizaw et al., 2010; Seid, 2017). The
majority of them are found in the lowland pastoral
and agro-pastoral production systems where
pastoralists in the South, East, and West keep them
mainly for milk and meat production purposes
(Gizaw et al., 2010). Goats reared in the mid- and
highlands are widely distributed in the crop-
livestock production systems with varying flock
sizes as a means of cash earnings and meat.
Despite the huge resource perspective, there is
limited information on the real genetic potentials of
local goat populations that are distributed in
various regions of the country. Phenotypic
J. Agric. Environ. Sci. Vol. 6 No. 1 (2021) ISSN: 2616-3721 (Online); 2616-3713 (Print)
Publication of College of Agriculture and Environmental Sciences, Bahir Dar University 20
characterization study is the basis for the
differentiation of indigenous animal genetic
resources and provides useful information in
designing appropriate genetic improvement
programs for their sustainable utilization and
conservation. Such studies mainly depend on the
knowledge of the variation of morphological traits,
which are important elements in the classification
of livestock based on appearance, size and shape
(Okpeku et al., 2011; Melesse et al., 2013). There
are different studies carried out by various scholars
to characterize the indigenous goat populations of
Ethiopia and most of them reported the existence of
phenotypic variations between and within the
studied goat ecotypes (Alemu, 2004; Hassen et al.,
2012; Gatew et al., 2015; Mekuriaw, 2016).
However, there is still limited characterization
studies conducted to describe the population
structure and genetic potentials of indigenous goat
populations found in Amhara Regional State in
general and that of South Gondar in particular. For
example, Hassen et al. (2012) conducted a
morphological characterization study on
indigenous goats in some zones of the Amhara
Region. Although South Gondar zone was
considered as one of them, the respective potential
districts for goat production were not adequately
represented due to limited samples included in that
particular study. As a result, the population
structure of the indigenous goats in South Gondar
zone has not been adequately described. Therefore,
this study was conducted to differentiate the
indigenous goat populations based on their
morphometric traits by applying the multivariate
analysis.
2. Materials and Methods
2.1. Site selection and sampling techniques
The study was conducted in three districts of South
Gondar zone in the Amhara Regional State,
Ethiopia. First, the relevant second-hand
information was gathered from the Agriculture and
Rural Development office of livestock. Based on
the information obtained, multi-stage purposive
sampling techniques were applied to select the
study districts and kebeles (the smallest
administrative unit within the district). In the first
stage, three districts namely Farta, Fogera and
Libokemkem were selected purposively based on
their goat population size and potential. In the
second stage, three kebeles from each district were
selected purposively based on the distribution of
the goat populations. In the third stage, the farmers
who own at least five or more matured goats of
both sexes were identified within kebeles.
Accordingly, 306 goats from Farta and Fogera and
the 204 goats from Lobokemkem districts with a
total number of 510 were sampled of which 153
and 357 were male and female goats, respectively.
The owner’s recall method along with dentition
classes (pairs of permanent incisors, PPI) was used
to estimate the ages of the goats. Accordingly,
goats with 1PPI, 2PPI, 3PPI and 4PPI were
classified in the age groups of yearling, 2-year-old,
3-year-old and 4-year-old and above, respectively
(Ebert and Solaiman, 2010). Each animal was
further identified by its sex and sampling site.
2.2. Types of data collected
Data on 14 morphometric traits were collected
according to the descriptor list of FAO (2012) for
phenotypic characterizations of goats. Accordingly,
the following traits were measured: live body
weight (LW), body length (BL), height at withers
(HW), heart girth (HG), chest depth (CD), chest
width (CW), paunch girth (PG), rump height (RH),
rump length (RL), rump width (RW), ear length
(EL), fore cannon circumference (FCC), teat length
(TL) and scrotal circumference (SC). All
morphometric traits were taken using plastic tape
while body weight using a suspended weighing
scale with 50 kg capacity by placing each animal in
self-devised holding equipment. All linear body
measurements were taken early in the morning
before goats leave for browsing.
2.3. Statistical Analysis
Since the sample size was unbalanced among
different districts and age groups, data were
subjected to GLM procedures by fitting district and
age as fixed effects. Since the interaction effect of
district by age was insignificant, it was dropped
from the analysis. The sex groups (males and
females) were analyzed separately due to
considerable differences in sample size. When F-
test declared significant, means were separated by
Least Square Means procedures. Accordingly, least
squares means with SE were presented rather than
the arithmetic mean. The procedure of the Cluster
analysis was performed and a dendrogram graph
was constructed based on Euclidean distance to
differentiate the goat populations of the three
districts using the average linkage method to group
the flocks into their morphological similarity.
Moreover, the stepwise discriminant analysis was
J. Agric. Environ. Sci. Vol. 6 No. 1 (2021) ISSN: 2616-3721 (Online); 2616-3713 (Print)
Publication of College of Agriculture and Environmental Sciences, Bahir Dar University 21
applied using the STEPDISC procedure to
determine which the morphometric variables have
more discriminating power. The relative
importance of the morphometric variables in
discriminating the three goat populations was
assessed using the level of significance, F-statistic
and partial R
2
. Collinearity among the variables
used in the discriminant model was also evaluated
using tolerance statistics. The canonical
discriminant analysis was then performed on the
identified variables with the most discriminating
power using the CANDISC procedure which
computed the Mahalanobis distances between class
means, uni- and multivariate statistics, and
canonical variables with eigenvalues. The
TEMPLATE and SGRENDER procedures were
also applied to create a plot of the first two
canonical variables in a scatter graph for visual
interpretation. Finally, discriminant analysis of the
DISCRIM procedure was conducted to determine
the percentage classification of goats into their
source populations using a quadratic discriminant
function for the unequal covariance matrices within
classes after checking with the Bartlett’s
homogeneity test. The classification accuracy of
the discriminant analysis was further cross-
validated by invoking the CROSSVALIDATE
option. All multivariate analyses were performed
using the Statistical Software of SAS (2012, ver.
9.4).
3. Results
3.1. Morphometric traits
In males, as shown in Table 1, the district showed
different influences across traits varying between
highly significant (LW, HG, CD, CW, RL, RW and
FCC) and non-significant (BL, HW, HR, EL and
SC). Fogera male goats had larger HG (p<0.05)
than those of Farta and Libokemkem while no
difference was noted between the latter two. Chest
width and CD values were larger (p<0.05) for Farta
and Fogera male goats than those of Libokemkem.
Male goats of Farta had larger (p<0.05) RL and
RW values than those of Fogera and Libokemkem
while those of Fogera had larger RL than those of
Libokemkem. A significantly bigger FCC was
observed in the male goat populations of Farta and
Libokemkem than those of Fogera.
The effect of the district in females was significant
for BL, HG, CD, PG, HR, and TL (Table 2).
Accordingly, the female goats of Libokemkem had
higher BL than those of Farta and Fogera. The
latter two did not significantly differ from each
other for the same trait. On the other hand, the
female goats of Fogera and Farta had larger HG
than those of Libokemkem. Likewise, the female
goats of Farta had larger CD and PG than those of
Fogera and Libokemkem. No significance
difference was noted between the latter two. A
significantly higher RH was observed in the female
goat populations of Fogera than those of
Libokemkem and had larger RL than those of the
Farta. No significant difference was found in RH
and RL values between Farta and Libokemkem
female goat populations. The female goats of Farta
had higher EL and TL than those of Libokemkem.
Teat length of goats from the Fogera district was
larger than that of Libokemkem. The effect of age
in both sexes was highly significant for all traits
except for EL showing a large heterogeneity among
males and females of different flocks.
J. Agric. Environ. Sci. Vol. 6 No. 1 (2021) ISSN: 2616-3721 (Online); 2616-3713 (Print)
Publication of College of Agriculture and Environmental Sciences, Bahir Dar University 22
Table 1: Least squares means of live weight (kg) and linear body measurement traits (cm) of indigenous male goat
populations as affected by district and age (N= 153)
Morphometric
traits
District
Overall
mean
(SEM)
Age
Overall
mean
(SEM)
Fixed effects
Farta
Libo-
kemkem
1PPI
2PPI
3PPI
4PPI
District
Age
Live weight
31.0
b
31.0
b
31.20.16
26.7
d
29.2
c
33.6
b
35.4
a
31.21.99
0.001
<.0001
Body length
62.4
62.2
62.30.23
58.4
d
60.4
c
64.5
b
66.1
a
62.41.78
0.576
<.0001
Height at
withers
69.7
69.8
69.80.13
66.3
d
68.6
c
71.4
b
72.8
a
69.81.45
0.434
<.0001
Heart girth
72.9
b
73.3
b
73.40.15
70.2
d
72.2
c
75.0
b
76.2
a
73.41.36
<.0001
<.0001
Chest depth
31.8
a
31.3
b
31.60.14
28.5
d
30.8
c
32.9
b
34.2
a
31.61.25
0.008
<.0001
Chest width
17.1
a
16.6
b
16.90.11
14.8
d
16.3
c
17.7
b
18.6
a
16.90.83
0.002
<.0001
Paunch girth
78.4
ab
78.1
b
78.50.34
74.0
d
77.2
c
80.3
b
82.8
a
78.61.91
0.014
<.0001
Height at
rump
72.0
72.2
72.10.17
68.8
d
71.2
c
73.5
b
74.9
a
72.11.34
0.821
<.0001
Rump length
16.7
a
15.4
c
16.00.13
14.9
d
15.6
c
16.3
b
17.4
a
16.10.53
<.0001
<.0001
Rump width
14.5
a
13.2
a
13.50.16
11.5
d
13.5
c
14.1
b
15.1
a
13.60.76
<.0001
<.0001
Fore canon
circumference
8.19
a
8.35
a
8.080.14
7.38
b
7.70
b
8.52
a
8.74
a
8.090.32
0.004
<.0001
Ear length
14.8
14.9
14.70.15
14.7
14.7
14.7
15.0
14.80.08
0.237
0.581
Scrotal
circumference
23.1
23.0
23.10.15
21.9
b
23.1
ab
23.7
a
23.9
a
23.20.45
0.332
<.0001
a,b,c,d
Means within district and age groups with different superscript letters are significant at p<0.05; SEM =
standard error of the mean
J. Agric. Environ. Sci. Vol. 6 No. 1 (2021) ISSN: 2616-3721 (Online); 2616-3713 (Print)
Publication of College of Agriculture and Environmental Sciences, Bahir Dar University 23
Table 2: Least squares means of live weight (kg) and linear body measurement traits (cm) of indigenous
female goat populations as affected by district and age (N = 357)
Morpho-
metric traits
District
Overall
mean
(SEM)
Age
Overall
mean
(SEM)
Fixed effects
Farta
Foge
ra
Libo-
kemkem
1PPI
2PPI
3PPI
4PPI
District
Age
Live weight
27.7
b
28.1
a
28.0
a
27.90.10
24.1
d
26.5
c
29.3
b
31.9
a
28.01.69
0.145
<.0001
Body length
60.2
a
60.1
a
59.5
b
59.90.11
56.6
d
58.7
c
60.8
b
63.7
a
60.01.52
<.0001
<.0001
Height at
withers
67.4
a
67.6
a
67.6
a
67.50.81
64.4
d
66.4
c
68.6
b
70.7
a
67.51.36
0.639
<.0001
Heart girth
70.7
b
71.2
a
71.3
a
71.00.10
68.0
d
70.1
c
72.2
b
73.8
a
71.01.26
0.015
<.0001
Chest depth
30.0
a
29.7
b
29.6
b
29.80.08
26.8
d
28.5
c
31.0
b
32.8
a
29.81.33
<.0001
<.0001
Chest width
15.4
15.4
15.4
15.40.07
13.7
d
15.1
c
16.0
b
16.8
a
15.40.66
0.283
<.0001
Paunch girth
75.6
a
74.7
b
74.4
b
74.90.16
70.6
d
72.7
c
77.3
b
79.0
a
74.91.96
<.0001
<.0001
Height at
rump
70.3
b
70.6
a
70.3
b
70.40.09
67.5
d
69.4
c
71.7
b
73.1
a
70.41.24
0.008
<.0001
Rump length
15.9
b
16.2
a
16.1
ab
16.00.06
15.1
d
15.7
c
16.4
b
17.1
a
16.10.43
0.051
<.0001
Rump width
13.2
a
12.8
a
13.1
a
13.00.12
11.0
c
13.2
b
13.9
a
14.0
a
13.00.70
0.236
<.0001
Fore canon
circumference
7.94
a
7.76
a
7.89
a
7.860.08
7.08
c
7.69
b
7.95
b
8.71
a
7.860.34
0.209
<.0001
Ear length
14.7
a
14.6
a
14.4
a
14.50.09
14.4
14.6
14.6
14.6
14.60.04
0.095
0.686
Teat length
3.71
a
3.72
a
3.66
a
3.700.02
3.04
3.49
4.02
4.24
3.700.27
0.007
<.0001
a,b,c,d
Means within district and age groups with different superscript letters are significant at p<0.05; SEM =
standard error of the mean
J. Agric. Environ. Sci. Vol. 6 No. 1 (2021) ISSN: 2616-3721 (Online); 2616-3713 (Print)
Publication of College of Agriculture and Environmental Sciences, Bahir Dar University 24
3.2. Multivariate Analysis
As indicated in Figure 1, the cluster analysis
generated a phylogenetic tree that clustered the
goat populations of South Gondar into two main
groups based on morphometric traits. The first
group included goat populations from Farta district
while the second group includes those from both
Fogera and Libokemkem districts as sub-cluster.
Table 3 presents the results of the stepwise
discriminant analysis showing Wilk’s Lambda, F-
values, probability and tolerance statistics. Twelve
quantitative traits were subjected to the STEPDISC
analysis of which eight were identified as the best
discriminating variables. Wilk’s lambda test
confirmed that all the selected variables had highly
significant (p<0.0001) contribution to differentiate
the total population into separate groups. The
variables with the highest discriminating power
were BL, HG, HW, PG, HR, RL, RW and FCC
(Table 3). The remaining four variables (LW, CD,
CW and EL) had poor discriminating power and
were excluded in the subsequent analysis.
All the eight variables were then subjected to
canonical discriminant analysis using the
CANDISC procedure. The univariate statistics
testing the hypothesis that class means are equal
which validate that each quantitative variable in
sampled populations is a significant (p<0.0001)
contributor to the total variation. The multivariate
statistics for differences between the districts was
also significant (p<0.0001). The hypotheses that
assumes the districts’ means are equal in the
populations was rejected by the Wilk’s Lambda
(p<0.0001) indicating that differences found
between studied districts were statistically different
from zero.
Figure 1: Dendrogram based on average linkage distance between goat populations reared in the three districts of
South Gondar using morphometric variables
Table 3: Summary of stepwise discriminant analysis for selection of traits
Step
Variables entered
Pr > F
Wilks'
Lambda
Pr<Lambda
ASCC
Pr>ASCC
Tolerance
1
Rump width
0.001
0.9743
0.0013
0.0129
0.0013
0.870
2
Heart girth
<.0001
0.9284
<.0001
0.0358
<.0001
0.687
3
Body length
<.0001
0.8679
<.0001
0.0672
<.0001
0.283
4
Puanch girth
0.003
0.8476
<.0001
0.0779
<.0001
0.239
5
Fore canon
circumference
0.004
0.8295
<.0001
0.0882
<.0001
0.233
6
Height at wither
0.113
0.8223
<.0001
0.0923
<.0001
0.129
7
Rump length
0.051
0.8126
<.0001
0.0977
<.0001
0.124
8
Height at rump
0.104
0.8053
<.0001
0.1019
<.0001
0.084
ASCC = average squared canonical correlation
J. Agric. Environ. Sci. Vol. 6 No. 1 (2021) ISSN: 2616-3721 (Online); 2616-3713 (Print)
Publication of College of Agriculture and Environmental Sciences, Bahir Dar University 25
The Mahalanobis distances and probability values
for the contrast among the goat populations of the
three districts are presented in Table 4. All
Mahalanobis distances were significant (p <
0.0001). The shortest distance (0.401) was
observed between the goat populations of Fogera
and Libokemkem districts while the larger between
those of Fogera and Farta (0.872). The
Mahalanobis distance between Farta and
Libokemkem goat populations was 0.858, which is
much similar to those of the Farta and Fogera.
Summary of canonical correlations and eigenvalues
are presented in Table 5. The multivariate statistics
for differences between the districts was significant
(p<0.0001). Standardized coefficients for the
canonical discriminant function, the canonical
correlation, the eigenvalues and share of total
variance accounted for this study revealed that both
canonical variables determined (CAN1 and CAN2)
was significant. The CAN1 and CAN2 accounted
for 70% and 30% of the total variation,
respectively. Table 5 further displayed the
likelihood ratio test of the Rao’s F approximation,
which rejected the hypothesis that assumes the
current canonical correlations and all smaller ones
are zero.
Figure 2 shows a plot built with the two canonical
variables illustrating the relationships between goat
populations belonging to different districts. The
plot displayed that CAN1 discriminates between
the two districts: Farta and Fogera while CAN2
best discriminates between Libokemkem and the
other two districts. However, it can be observed in
the figure that there is a visible overlapping among
the three goat populations indicating the existence
of homogeneity.
Table 4: Pairwise Mahalanobis distance values among
the three goat populations
Districts
Farta
Fogera
Libokemkem
Farta
0
0.872
0.858
Fogera
0.872
0
0.401
Libokemkem
0.858
0.401
0
All distances are significant at p<0.001
Table 5: Summary of canonical correlations, eigenvalues and likelihood ratios
Functions
Canonical
correlations
Eigenvalues
Likelihood
ratio
Approximate
F-value
Pr>F
Eigen-value
Proportion
Cumulative
CAN1
0.373
0.161
0.70
0.700
0.805
7.15
<0.0001
CAN2
0.254
0.069
0.30
1.000
0.935
4.96
<0.0001
CAN1 = canonical variable 1; CAN2 = canonical variable 2
Figure 2: Canonical representation of the goat population in the three districts of South Gondar
Discriminant analysis assumes that the individual
group covariance matrices are equal (homogeneity
within covariance matrices) and by default, it uses
the linear discriminant function for classification.
J. Agric. Environ. Sci. Vol. 6 No. 1 (2021) ISSN: 2616-3721 (Online); 2616-3713 (Print)
Publication of College of Agriculture and Environmental Sciences, Bahir Dar University 26
In the current discriminant analysis, equality of
covariance matrices within groups was tested using
Bartlett’s test of homogeneity for all traits and was
significant (
2
= 248; p<0.0001) by rejecting the
null hypothesis that assumes all covariance
matrices within the goat populations are
homogenous. Therefore, the within-group
covariance matrices were used to derive the
quadratic discriminant function criterion for the
classification of the three goat populations.
As presented in Table 6, the discriminant analysis
correctly classified 58.6% of Farta goats into their
respective origin population with 22.4 and 19.1%
being misclassified to Fogera and Libokemkem
goat populations, respectively. Similarly, 62.3% of
Fogera goats were correctly assigned to their
source population while the remaining 14.9 and
22.7 being misclassified to Farta and Libokemkem
goat populations, respectively. The quadratic
discriminant function further differentiated the
Libokemkem goat from others with 63.2% correct
classification into their original source population
with the remaining 13.7 and 23.0% being
misclassified to Farta and Fogera goat populations,
respectively. The overall error count estimates
provide the proportion of misclassified
observations in each group being highest in Farta
goats (41.4%) and lowest in those of Libokemkem
(36.8%).
The classification accuracy of the discriminant
analysis was further cross-validated and indicated
an overall 53.0% success rate (Table 6). The error
count estimate for the cross-validation was 46.7,
50.7, and 43.6% for Farta, Fogera and
Libokemkem districts, respectively. The overall
error count estimate for classification was 38.6%
while it was 47.0% for cross-validation option. It
would be worthwhile to note that the cross-
validation option achieved a fairly unbiased
estimate with a relatively large variance.
Table 6: Percent of individual goats classified into their respective districts and cross-validation of classification
based on morphometric variables (values in brackets are number of goats)
Districts
Farta
Fogera
Libokemkem
Total
Re-substitution
Farta
58.6 (89)
22.4 (34)
19.1 (29)
100 (152)
Fogera
14.9 (23)
62.4 (96)
22.7 (35)
100 (154)
Libokemkem
13.7 (28)
23.0 (47)
63.2 (129)
100 (204)
Error count estimate
0.414
0.377
0.368
0.386
Priors
0.333
0.333
0.333
-
Cross-validation
Farta
53.3 (81)
25.0 (38)
21.7 (33)
100 (152)
Fogera
18.2 (28)
49.4 (76)
32.5 (50)
100 (154)
Libokemkem
16.2 (33)
27.5 (56)
56.4 (115)
100 (204)
Error count estimate
0.467
0.507
0.436
0.47
4. Discussion
4.1. Quantitative traits
Height at withers and EL were not affected by the
district in both sex groups which indicates their
similarity in their heights and the size of the ears.
In male goats, the district did not affect BL, HR
and SC. Similarly, LW, CW, RL and RW in
females were not affected by the district. These
observations suggest that the genetic improvement
of these traits in both sexes of the studied districts
might be justifiable. Age of the goats had a
significant (P<0.05) effect on all linear body
measurement traits except for EL in both sexes.
Most of the morphometric traits increased with age,
which is in close agreement with the findings of
Lorato et al. (2015) for Woyto Guji goat type in
Loma district. A linear increment of morphometric
traits with age indicates a normal body
development of goats which suggest the suitability
of the production environment for goat rearing. The
absence of an increase in the size of ear in both
sexes may suggest its limited importance during the
physical development of the goats.
In the present study, male goats appeared to be
heavier than females. This might be explained by
the presence of a high level of the hormone
androgen in bucks, which is responsible for
muscular growth. This observation was in
agreement with that of Seid et al. (2016) who
reported similar findings in goats reared in the
Western highland of Wollega zone, Oromia. On the
J. Agric. Environ. Sci. Vol. 6 No. 1 (2021) ISSN: 2616-3721 (Online); 2616-3713 (Print)
Publication of College of Agriculture and Environmental Sciences, Bahir Dar University 27
contrary, Jeda and Asefa (2016) reported that
females goats raised in the Bale zone had higher
LW than those of bucks. Such inconsistencies
might be due to differences in age, breed,
management, and accuracy of taking measurements
in which the data were collected. Moreover, such
differences might be attributed to the result of
negative selection practiced by the farmers as fast-
growing male kids are sold at an earlier age.
Female goats had higher RL than males and FCC
and RW values almost the same for both sexes.
Rump size is a very important structure for ease
deliverance.
The SC is an important trait that is closely
associated with the testicular growth and sperm
production capacity of domestic animals. Gatew et
al. (2015) reported relatively higher SC values for
bucks in eastern Ethiopia than observed in the
current study (27 vs. 23 cm). Since SC size is
dependent on the maturity of the animal, the
differences could be attributed to the age of bucks
when data were collected. This has been supported
by the present observation in which SC has been
significantly affected by age. Moreover, SC
showed a significant positive correlation with the
live weight of goats (Tade et al., unpublished data),
which substantiated the dependency of SC on the
body development of the animal. Consistent with
the current results, Raji and Ajala (2015) observed
a significant effect of body weight on SC for West
African Dwarf buck. As SC is an indirect
measurement of testicular size, knowing the
increased size of testis may be used as an indicator
in the onset of active spermatogenesis and, hence,
the possibility of using bucks for breeding at an
earlier age than normally recommended. Such
knowledge might be particularly essential if young
bucks may not be kept together with the does for
reasons related to control of the occurrence
concurrence of inbreeding as well as disease
transmission during mating.
The teat length positively affected milk production
potentials of does (El-Gendy et al., 2014). Merkhan
and Alkass (2011) reported 3.6 cm TL for Iraqi
Black and Meriz goats, which is comparable with
the current findings (3.7 cm). The overall teat
length of goats in the present study is
comparatively higher than reported by Alemayehu
et al. (2015) for goats of West Amhara (3.7 vs.
3.40 cm). Teat length is significantly affected by
the age of the does, which indicates a linear
increase in teat size with the advancing age. In
general, the goat populations of Fogera district had
the highest BW and HG values as compared to
those of Farta and Libo-kemkem districts.
However, the goats of the Farta district had the
highest CD. The observed variations might
associated with differences in the management
practices among the communities and the
availability of feed and water resources.
4.2. Multivariate Analysis
The Cluster analysis classified the goat populations
in two main groups based on their morphometric
traits in which the first group included the Farta
goat populations while the second group includes
the Fogera and Libokemkem goats as sub-cluster.
This observation indicates that goats of the Fogera
and Libokemkem districts are much closer to each
other than those of the Farta and confirmed the
results of the cross-classification of population
distribution with discriminant analysis.
Except for height at rump, the tolerance values
obtained from the present study were greater than
0.1. This is an indication that there was no
collinearity problem among the eight most
discriminating morphometric variables (Yakubu et
al., 2010; Selolo et al., 2015). Some of the present
discriminant variables are similar to those reported
by Yakubu et al. (2011) for West African Dwarf
and Red Sokoto goat breeds.
All Mahalanobis distances were significant which
are in line with the findings of Zaitoun et al.
(2005). The Mahalanobis distance was relatively
short for the goat populations of Fogera and
Libokemkem districts, indicating that they are
homogenous in their morphometric characters
probably due to sharing similar genetic identities.
This trend has been demonstrated in the
dendrogram displayed in Fig. 1. The low
Mahalanobis distances between Fogera and
Libokemkem goat populations might have resulted
from non-selection, continuous inbreeding, and
high intermingling rate between these two
populations in an open management system of
production, which is commonly practiced by many
rural communities. Moreover, the two districts
(Fogera and Libokemkem) are sharing quite a
substantial area of borderline which affirms the
homogeneity of the genetic identity of the goats
resulting from intermix of genetic materials. On the
other hand, the Mahalanobis distance was
J. Agric. Environ. Sci. Vol. 6 No. 1 (2021) ISSN: 2616-3721 (Online); 2616-3713 (Print)
Publication of College of Agriculture and Environmental Sciences, Bahir Dar University 28
comparatively larger between goat populations of
Farta and Libokemkem as well as between those of
Farta and Fogera districts. Comparative large
Mahalanobis distances were reported between
Damascus and Dhaiwi goat breeds while a
moderate distance was observed between Mountain
and Dhaiwi (Zaitoun et al., 2005). In general, the
Mahalanobis distances between the three goat
populations were too short indicating the existence
of homogeneity among the studied goat
populations.
The univariate statistics testing the hypothesis that
class means are equal shows that each quantitative
variable in sample populations except HG, HW and
HR was significant (p<0.05) contributor to the total
variation. The Wilks’ lambda, the ratio of within-
group variability to total variability on the
discriminator variables, is an inverse measure of
the importance of the discriminant functions. The
Wilks’ lambda test (Table 5) for the population was
0.805, which reflects less part (19.5%) of the
variability in the discriminator variables was
because of the differences between populations
rather than variation within the population.
In the current study, CAN1 and CAN2 accounted
for 70% and 30% of the total variation,
respectively, indicating a complete representation
of individuals of the local goat populations with
one scatter plane. The extracted both canonical
variables were found to be significantly different,
which agrees with the observations of Traoré et al.
(2008) for goat populations of Burkina Faso. On
the contrary, Selolo et al. (2015) reported that the
CAN1 was significant while CAN2 remained
insignificant for local South African goats. There
are indeed conflicting reports in the literature on
the proportion of total variation explained by both
canonical variables (CAN1 and CAN2). For
example, Traoré et al. (2008) reported a total
variation of 94.0% and 5.5% for CAN1 and CAN2,
respectively while the corresponding values were
82.4% and 10.7% for Jordan native goat breeds as
reported by Zaitoun et al. (2005). Selolo et al.
(2015) reported that 91.9% of the total variation
was accounted by CAN1 while only 8.1% by
CAN2. The reported differences in the literature
might be explained by the sample size, age and
breed of goats studied. Rump width, PG, RL and
BL dominated CAN1, while FCC showed the
largest influence on CAN2. Herrera et al. (1996)
found that head length and withers height were the
most important variables in CAN1, while head
width and shin circumference were the most
important variables in CAN2 in their
discrimination among the five Andalusian goat
breeds.
The values computed for CAN1 and CAN2 for
each individual were plotted by districts and
displayed in Figure 2. Accordingly, the Farta
individuals appeared to relatively homogeneous
and clustered together on the right hand of the X-
axis; the Libokemkem are mainly distributed on the
positive values of the Y-axis, and the Fogera
individuals showed an intermediate distribution but
inclined toward to the Libokemkem goats. In this
respect, the discriminant analysis carried out
provided complementary information (Table 6) in
which most of the goat populations of Libokemkem
and Fogera districts were classified into their
source population (63.2% and 62.3%, respectively)
whilst the rest (13.7 and 14.9) were misclassified as
Farta individuals. The discriminant analysis also
classified 58.6% Farta individual indigenous goats
into their original districts. Similarly, Selolo et al.
(2015) found that 60.3, 58.1 and 38.5% of the
individual goats were classified into their original
agro-ecological zones with several individuals
being misclassified. Yakubu and Ibrahim (2011)
reported that Yankasa (45.9%), Uda (33.5%) and
Balami (61.5%) sheep breeds were correctly
classified into their source group. Another study
conducted by Dossa et al. (2007) indicated that
more than 70% of individual goats were correctly
allocated to their different source groups. Similarly,
the respective 79.3% and 82.7% of Sudan and
Sudan-Sahel goat populations of Burkina Faso
were classified into their source population (Traoré
et al., 2008). Dekhili et al. (2013) reported that
73%, 66.8% and 79.3% of Algeria goats were
classified into North, Center and South
environments, respectively. Yakubu et al. (2010)
reported that only 24.4% of rainforest and 22.9% of
guinea savanna goats were correctly classified into
their source populations. These reports suggest the
importance of multivariate discriminant analysis to
differentiate indigenous livestock populations that
are being reared in various production
environments.
Based on the discriminant analysis, the overall
average error count estimate was 38.6% for all
observations and 61.5% of the overall sampled
populations were correctly classified to their origin
J. Agric. Environ. Sci. Vol. 6 No. 1 (2021) ISSN: 2616-3721 (Online); 2616-3713 (Print)
Publication of College of Agriculture and Environmental Sciences, Bahir Dar University 29
districts indicating the heterogeneity of goat
populations within districts for those variables
included in the discriminant analysis. The relatively
higher errors of classification for Farta goats may
indicate that they might have been extensively
mixed with the other local goat populations. The
misclassification observed in this study may also
suggest that the level of genetic exchange that has
taken place overtime between the goats reared in
the three districts. In addition, there is a good
possibility of admixture among these goats because
of the continuous migration of flocks that existed
for many generations. Moreover, the assignment
errors of local goat breeds might have occurred
between goat populations reared in the same
production system. Such speculations might be
justifiable, as goats reared under the same
production system might have been selected
naturally and artificially for similar traits (Zaitoun
et al., 2005).
The low differentiation assessed between Farta and
Fogera and between Farta and Libokemkem goats
using the Mahalanobis distances and the large
classification errors obtained using the discriminant
analysis did not give statistical support to separate
these goat populations into different distinct
ecotypes. Moreover, a significant proportion of
cross-classification errors (41.5%) observed in
Farta goat populations suggests that they might
share a similar genetic basis with the other two goat
populations. Such admixtures are possible due to
the existence of an active marketing system of
goats in the region.
5. Conclusion
Bucks of Fogera district had larger heart girth than
those of Farta and Libokemkem while the females
of Libokemkem district had higher body length
than those of Farta and Fogera. The cluster analysis
showed two separate clusters: cluster one included
the Farta goat populations as one distinct group
while cluster two included the Fogera and
Libokemkem goats under one sub-cluster. The
canonical discriminant analysis verified similar
trend by indicating the Fogera and Libokemkem
goats are the closest while the largest Mahalanobis
distance was between Farta and Fogera goat
populations with all distances being significant.
However, the canonical discriminant analysis
indicated a visible overlapping among the three
goat populations suggesting the existence of
homogeneity. The respective 58.6%, 62.3% and
63.2% of the Farta, Fogera and Libokemkem goat
populations were correctly classified into their
districts with overall error count estimates of
38.6%. The accuracy of the classification was
further cross-validated in which the respective
53.3, 49.4 and 56.4 of Farta, Fogera and
Libokemkem goats were correctly assigned to their
source populations with an overall error count of
47.0%.
Conflict of Interest
The authors declare that there is no conflict of
interest.
Acknowledgements
The authors are highly grateful to the Agarfa
Agricultural Technical and Vocational Educational
Training (ATVET) College (Ethiopia) for
supporting this research work. The first author
would like to thank the Agarfa ATVET College for
granting him study leave with all benefits. Finally,
we are also grateful to individual households who
have fully collaborated with us to take all the
morphometric measurements from their animals for
this study.
References
Alemayehu, K., Kebede, D., Melese, A., and
Andualem, S. (2015). Selection of indigenous
goat types for designing product specific
breeding strategy in selected districts of West
Amhara, Ethiopia. World Applied Sciences
Journal 33 (8): 1279-1285.
Alemu, T. (2004). Genetic characterization of
indigenous goat populations of Ethiopia using
microsatellite DNA markers. PhD.
Dissertation, National Dairy Research
University, India.
Arandas, J.K.G., da Silva, N.M.V., Nascimento,
R.B., Filho E.C.P., Brasil L.H.A., and Ribeiro
M.R. (2017). Multivariate analysis as a tool for
phenotypic characterization of an endangered
breed. Journal of Applied Animal Research
45(1): 152-158 https://doi.org/10.1080/ 09712
119.2015.1125353
CSA (Central Statistical Agency), 2018/19. Federal
democratic republic of Ethiopia Central
statistical agency Agricultural sample survey.
Volume II Report on livestock and livestock
characteristics.
https://www.statsethiopia.gov.et/our-survey-
reports/.
J. Agric. Environ. Sci. Vol. 6 No. 1 (2021) ISSN: 2616-3721 (Online); 2616-3713 (Print)
Publication of College of Agriculture and Environmental Sciences, Bahir Dar University 30
Dekhili, M., Bounechada, M., and Mannalah, I.
(2013). Multivariate analyses of morphological
traits in Algerian goats, Sétif, north-eastern
Algeria. Animal Genetic Resources 52: 51-57.
Dossa, L.H., Wollny, C., and Gauly, M., 2007.
Spatial variation in goat populations from
Benin as revealed by multivariate analysis of
morphological traits. Small Ruminant
Research 73: 150-159.
Ebert, R.A., and Solaiman, S.G. (2010). Animal
Evaluation. In Solaiman, S.G. 2010 (ed.), Goat
Science and Production (pp 77-78). Iowa,
USA. Blackwell Publishing, John Wiley &
Sons, Inc. Publication.
El-Gendy, M.E., Youssef, H.F., Saifelnasr, E.O.H.,
El-Sanafawy, H.A., and Saba, F.E. (2014).
Relationship between udder characteristics and
each of reproductive performance and milk
production and milk composition in Zaraibi
and Damascus dairy goats. Egyptian Journal of
Sheep and Goat Sciences 9(3): 95-104.
FAO (Food and Agricultural Organization of UN).
(2012). Phenotypic characterization of animal
genetic resources. FAO Animal Production
and Health Guidelines No.11. Rome, Italy.
Gatew, H., Hassen, H., Kebede, K., and Haile, A.
(2015). Characterization of indigenous goat
populations in selected areas of Ethiopia.
American-Eurasian Journal of Scientific
Research 10 (5): 287-298.
Gizaw, S., Tegegne, A., Gebremedhin, B., and
Hoekstra, D. (2010). Sheep and goat
production and marketing systems in Ethiopia:
Characteristics and strategies for improvement.
IPMS (Improving Productivity and Market
Success) of Ethiopian Farmers Project
Working Paper 23. ILRI (International
Livestock Research Institute), Nairobi, Kenya.
Hassen, H., Baum, M., Rischkowsky, B., and
Tibbo, M. (2012). Phenotypic characterization
of Ethiopian indigenous goat- populations.
African Journal of Biotechnology, 11(73),
13838-13846.
Herrera, M., Rodero, E., Gutierrez, M. J., Pena, F.,
and Rodero, J. M. (1996). Application of
multifactorial discriminant analysis in the
morphostructural differentiation of Andalusian
caprine breeds. Small Ruminant Research 22:
39-47.
Jeda, G., and Asefa, B. (2016). Characterization of
indigenous goat’s type using morphological
character and body measurements in Sinana
district, Bale zone, South East Ethiopia. Global
Journal of Science Frontier Research: D
Agriculture and Veterinary 16(7): 1-8.
Lorato, Y., Manzoor, K., and Belay, B. (2015).
Morphological characterization of indigenous
Woyto Guji goat type in Loma district,
Southern Ethiopia. African Journal of
Agricultural research 10(20): 2141-2151.
Mekuriaw, M. (2016). Molecular characterization
of Ethiopia indigenous goat population;
genetic diversity and structure, demographic
dynamics and assessment of the kisspeptin
gene polymorphism. PhD. Dissertation. Addis
Ababa University, Addis Ababa, Ethiopia.
Melesse, A., Banerjee, S., Lakew, A., Mersha, F.,
Hailemariam, F., Tsegaye, S., and Makebo, T.
(2013). Morphological characterization of
indigenous sheep in Southern Regional State,
Ethiopia. Animal Genetic Resources: 52: 39-
50.
Merkhan, K.Y., and Alkass, J.E. (2011). Influence
of udder and teat size on milk yield in Black
and Meriz goats. Research Opinions in Animal
and Veterinary Sciences 1(9): 601-605.
Okpeku, M., Yakubu, A., Peters, S., Ozoje, M.,
Ikeobi, C., Adebambo, O., and Imumorin, I.
(2011). Application of multivariate principal
component analysis to morphological
characterization of indigenous goats in
southern Nigeria. Acta Argiculturae Slovenica
98 (2): 101-109.
Peacock, C. (2005). Goats: A pathway out of
poverty. Small Ruminant Research 60(1): 179-
186.
Raji, L.O., and Ajala, O.O. (2015). Scrotal
circumference as a parameter of breeding age
for West African Dwarf bucks. Turkish Journal
of Agriculture - Food Science and Technology
3(8): 668-671.
SAS (Statistical Analysis System). (2012). SAS for
Windows, ver. 9.4 SAS Institute, Inc., Cary,
NC, USA.
Seid, A., Kebede, K., and Effa, K. (2016).
Morphological characterization of indigenous
goats in Western Ethiopia: implication for
community-based breeding programs. Animal
Genetic Resources 58: 53-62.
Seid, A. (2017). Breeding practices and strategies
for genetic improvement of indigenous goats
in Ethiopia. Greener Journal of Agricultural
Sciences 7 (4): 90-96.
J. Agric. Environ. Sci. Vol. 6 No. 1 (2021) ISSN: 2616-3721 (Online); 2616-3713 (Print)
Publication of College of Agriculture and Environmental Sciences, Bahir Dar University 31
Selolo, T.C., Majela, L.M., Norris, D., Ng’ambi,
J.W., and Brown, D. (2015). Morphological
differentiation of indigenous goats in different
agro-ecological zones of Vhembe district,
Limpopo province, South Africa. Indian
Journal of Animal Research 49 (4): 527-531.
Traoré, A., Tamboura, H.H., Kaboré, A., Royo,
L.I., Fernández, I., Álvarez, I., Sangaré, M.,
Bouchel, D., Poivey, J.P., Francois, D.,
Sawadogo, L., and Goyache, F. (2008).
Multivariate analyses on morphological traits
of goats in Burkina Faso. Archiv Tierzucht
51(6): 588-600.
Yakubu, A., Salako, A.E., Imumorin, I.G., Ige,
A.O., and Akinyemi, M.O. (2010).
Discriminant analysis of morphometric
differentiation in the West African Dwarf and
Red Sokoto goats. South African Journal of
Animal Science 40: 381-387.
Yakubu, A., and Ibrahim, I.A. (2011). Multivariate
analysis of morphostructural characteristics in
Nigerian Indigenous sheep. Italian Journal of
Animal Science 10: 83-86.
Yakubu, A., Salako, A.E., and Imumorin, I.G.
(2011). Comparative multivariate analysis of
biometric traits of West African Dwarf and
Red Sokoto goats. Tropical Animal Health and
Production 43: 561-566.
Zaitoun, I.S., Tabbaa, M.J., and Bdour, S. (2005).
Differentiation of native goat breeds of
Jordanian the basis of morphostructural
characteristics. Small Ruminant Research 56:
173-182