Full Length Research Article

Science World Journal Vol 6 (No 2) 2011

www.scienceworldjournal.org
ISSN 1597-6343

PRINCIPAL COMPONENT ANALYSIS OF

BODY MEASUREMENTS IN THREE STRAINS
OF BROILER CHICKEN.
* UDEH. I1 & OGBU. C. C2
those components that define body conformation. These could be
1Department of Animal Science, Delta State University, Asaba used as selection criteria for improving meatiness in broilers.
Campus.
2Department of Animal Science, University of Nigeria, Nsukka. MATERIALS AND METHODS
*drudeh2005@yahoo.com Study location: The study was conducted at the poultry unit of
teaching and research farm, Department of Animal Science, Delta
ABSTRACT State University, Asaba campus. The Delta State University Asaba
This study was conducted to explore the relationship among body campus is located between longitude 60º 45¹ East and latitude 60º
measurements in 3 strains of broilers chicken (Arbor Acre, Marshal 12¹ North.
and Ross) using principal component analysis with the view of
identifying those components that define body conformation in
Experimental animals and their management: A total of 210
broilers. A total of 180 birds were used, 60 per strain. The parameters
recorded at 8 weeks of age were body weight, shank length, thigh broiler chicks comprising 70 each of Arbor Acre, Marshal and Ross
length, drumstick length, body length, body width, breast width and procured from a reputable hatchery were used for the study. The
wing length. Principal component analysis with variance maximizing actual data collection was done using 60 broilers per strain while
orthogonal rotation was used to extract the components. Two 10 chicks were provided to take care of mortality. The chicks were
principal components were extracted in Arbor Acre which explained housed separately in deep litter pens at day old. The birds were
65% of the total variation in the original variables. Similarly three fed ad libitum top broiler starter diet from day old to 4 weeks of age
principal components each were extracted in Marshal and Ross and a top feed broiler finisher diet from 4 to 8 weeks of age. Clean
strains accounting for 74.76% and 70% of the total variance
drinking water was also made available to the birds all the time. All
respectively. Generally, PC1 had the largest share of the total
variance and correlated highly with breast width, wing length, thigh the necessary vaccines for broiler chicks were administered at the
length, shank length and body length. PC1 was termed the appropriate ages.
generalized form of broilers. PC2 had its loadings on drum stick
length, breast width, thigh length and wing length while PC3 had Traits measured: The body weights of the birds were recorded on
positive loading on drumstick length. These components could be weekly basis to 8 weeks of age. The body measurements namely
used as selection criteria for improving body size of broilers. shank length, thigh length, drumstick length, body length, body
width, breast width and wing length were measured at 8 weeks of
Key words: Body parameters, Broilers, Loadings, Variance
age as described by Monsi (1992) and Udeh et al., (2011).
maximizing rotation.
INTRODUCTION Statistical analysis: Means, standard errors, minimum and
Growth is a complex trait in animals that is controlled by genetic maximum of body weight and body measurements of each strain
and non genetic factors. Body weight and body conformations are of broilers were obtained using the descriptive statistic of SPSS 16
the two important parameters for measuring growth in the (2007). One way analysis of variance was used to test the effect of
domestic chicken. The mechanisms involved in the control of strain of broiler on the parameters. Pearson correlation coefficients
growth in chickens are too complex to be explained only under among the body measurements were calculated for each strain of
univariate analysis because all related traits are biologically broilers and the correlation matrix which was the primary data
correlated due to pleiotropic effect of genes and linkage of loci required for PCA generated. Bartletts test of sphericity was used to
(Rosario et al., 2003). Consequently, multivariate approach is test if the correlation matrix was an identity matrix (each variable
employed to analyze growth data in chicken and other domestic correlated with itself) or a correlation matrix full of zero. The
animals. Principal components analysis is a mathematical suitability of the data set to PCA was further tested by Kaiser-
procedure that transforms a number of possibly correlated Meyer-Olkin (KMO) measure of sampling adequacy. This tested
variables into a smaller number of uncorrelated variables known as whether the partial correlations among variables were small. A
principal components which are ordered so that the first few retain KMO measure of 0.60 and above is considered adequate
most of the variation present in the original variables (Jolliffe, (Eyduran et al., 2010). Everitt et al., (2001) defined principal
2002). From the view point of animal genetics and improvement, component analysis as a method of transforming variables in a
principal components simultaneously consider a group of attributes multivariate data set, x1, x2, ……..xp into new uncorrelated
which may be used for selection purpose (Pinto et al., 2006). variables y1, y2,……….yp which account for decreasing proportions
Principal component analysis has been used to describe the of the total variance in the original variables defined as
relationship between body measurements and body size in
chicken (Ibe, 1989; Yakubu et al., 2009a), duck (Ogah et al., 2009) y1=a11x1+a12x2+ ……+ a1pxp.
and turkey (Ogah et al., 2011). Arbor Acre, Marshal and Ross y2=a21x1+a22x2+……+ a2pxp.
were among the strains of broiler chicken reared by farmers in yp=ap1x1+ap2x2+..….+ appxp.
southern Nigeria. They cope fairly well with the hot season of
January to March in Nigeria and reach market weight at about 8 The principal components y1, y2…….yp account for decreasing
weeks of age. Understanding the interrelationships between body proportions of the total variance in the original variables x1,
weight and body measurements in these birds will help the farmer x2,……xp. Variance maximizing orthogonal rotation was used in the
predict their body weight at various ages especially in the rural linear transformation of the factor pattern matrix in order to make
areas where scales may not be available. The objective of this the interpretation of the extracted principal components easier.
study was to examine the relationship among body measurements The principal components analyses were performed using the
in the three strains of broiler chickens with a view of identifying factor program of SPSS 16 (2007) statistical package.

11

Principal Component Analysis of Body Measurement in Three Strains of Chicken.

Science World Journal Vol 6 (No 2) 2011
www.scienceworldjournal.org
ISSN 1597-6343

RESULTS
Table 1 presents the descriptive statistics of body weight and body length and wing length compared to Ross and Marshal strains.
measurements of three strains of broiler chicken at 8 weeks of The Ross strain was significantly (p<0.05) superior to Arbor Acre
age. Arbor Acre and Ross attained average body weight of 1.88 kg and Marshal strain in body length while Marshal strain recorded
and 1.81 kg respectively which were superior to Marshal (1.65kg) the highest breast width. The three strains of broilers did not differ
at 8 weeks of age. Arbor Acre was the most superior in drumstick (p>0.05) in shank length and body width.

TABLE 1. DESCRIPTIVE STATISTICS OF BODYWEIGHT (KG) AND LINEAR BODY MEASUREMENTS (CM) OF
THREE STRAINS OF BROILER CHICKEN AT 8 WEEKS OF AGE.

Arbor Acre Marshal Ross

Variable No
Mean ± S.E Min Max Meant ± S.E Min Max Mean ± S.E Min Max
Bodyweight 60 1.88 ± 0.04b 1.60 2.20 1.65 ± 0.03a 1.40 1.90 1.81 ± 0.03b 1.60 2.00
Shanklength 60 7.80 ± 0.05a 7.10 8.20 7.64 ± 0.06a 7.10 8.00 7.54 ± 0.06a 7.10 8.00
Thighlength 60 17.16 ±0.06b 16.70 17.60 17.39 ± 0.08b 16.80 17.90 15.25± 0.04a 15.00 15.60
Drumstick length 60 8.72 ± 0.03c 8.50 9.00 7.95 ± 0.05b 7.50 8.40 7.46 ± 0.07a 6.90 7.90
Bodylength 60 32.79 ± 0.15b 31.40 33.80 31.30 ± 0.16a 30.50 32.50 37.12 ± 0.07c 36.70 37.70
Bodywidth 60 16.48 ± 0.05a 15.90 16.90 16.45 ± 0.10a 15.20 17.10 16.67 ± 0.06a 16.20 17.20
Breastwidth 60 10.53 ± 0.05b 9.90 10.90 12.93 ±0.05c 12.60 13.40 9.60 ± 0.06a 8.90 9.90
Winglength 60 18.98 ± 0.07c 18.30 19.50 17.95 ± 0.06a 17.50 18.50 18.57 ± 0.05b 18.10 19.00
Mean in the same row bearing different superscript letters are significantly (p<0.05) different.

Table 2 presents the coefficient of correlations of body weight and Similarly, in Ross strain, significant (p<0.05) positive relationships
body measurements of the three strains of broiler chicken. The were observed for thigh length and wing length (0.54), shank
correlation coefficients ranged -0.05-0.76, -0.02-0.56 and -0.28- length and wing length (0.55) and body width and breast width
0.60 in Arbor Acre, Marshal and Ross strains respectively. (0.50). Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy
Relationships between body weight and most of the body computed for Arbor Acre, Marshal and Ross were 0.60, 0.62 and
measurements were positive and non significant (p>0.05) in the 0.67 respectively. These values revealed the pattern of correlation
three strains of broilers. Highly significant (p<0.01) positive among the variables. Though not marvelous, the values indicated
correlations were recorded for shank length and breast width that the sample sizes were adequate to apply PCA (Kaiser, 1960).
(0.62), thigh length and breast width (0.66), thigh length and wing Results of the Bartlett test of sphericity for body measurements of
length (0.64) and breast width and wing length (0.76) in Arbor Acre Arbor Acre (chi-square 55.70, p=0.00), Marshal (chi-square 37.71,
broiler. In Marshal, significant (p<0.05) positive relationships were p<0.01) and Ross (chi-square 64.97, p=0.00) were significant. This
obtained for shank length and body width (0.56), shank length and implied that PCA was applicable to the data sets.
wing length (0.50) and body width and wing length (0.47).

TABLE 2. CORRELATION COEFFICIENT OF BODYWEIGHT AND BODY MEASUREMENTS OF ARBOR ACRE,

MARSHAL AND ROSS BROILER CHICKEN AT 8 WEEKS OF AGE.

TRAIT BW SHL THL DSL BDL BDW BRW WL

Arbor Acre.
BW 1.00
SHL 0.19 1.00
THL 0.11 0.31 1.00
DSL -0.05 0.11 -0.01 1.00
BDL -0.43 0.12 0.31 0.34 1.00
BDW 0.10 -0.08 0.26 0.45* 0.09 1.00
BRW 0.26 0.62** 0.66** -031 -0.04 0.09 1.00
WL 0.18 0.49* 0.64** -0.30 0.09 0.12 0.76** 1.00

Marshal
BW 1.00
SHL -0.02 1.00
THL 0.21 0.31 1.00
DSL 0.27 -0.53* -0.52* 1.00
BDL 0.50* 0.01 0.20 0.18 1.00
BDW -0.17 0.56* 0.16 -0.31 -0.34 1.00
BRW -0.31 0.20 -0.42 0.05 -0.15 -0.04 1.00
WL -0.15 0.50* 0.19 -049* -0.05 0.47* 0.03 1.00

Ross
BW 1.00
SHL 0.36 1.00
THL -0.40 0.22 1.00
DSL 0.39 -0.22 -0.23 1.00
BDL -0.37 -0.28 0.35 -0.01 1.00
BDW 0.03 0.17 -0.46* -0.42 -0.70** 1.00
BRW 0.60** 0.44 -0.48* 0.23 -0.71** 0.50* 1.00
WL 0.17 0.55* 0.54* -0.37 0.09 -0.08 -0.06 1.00

*Significant (P<0.05) **Significant (p<0.01)

12

Principal Component Analysis of Body Measurement in Three Strains of Chicken.

Science World Journal Vol 6 (No 2) 2011
www.scienceworldjournal.org
ISSN 1597-6343

Table 3 presents the eigen values, percentage of the total variance accounted for by the components. It ranged 0.413-0.883, 0.592-
along with the rotated component matrix and communalities of the 0.894 and 0.760-0.940 in Arbor Acre, Marshal and Ross broilers
body measurements of three strains of broiler chicken. The respectively. The eigen values showed the amount of variance out
communalities represent estimates of the variance in each variable of the total variance explained by each of the factors.

TABLE 3. EIGEN VALUES AND PERCENTAGE OF TOTAL VARIANCE ALONG WITH THE ROTATED COMPONENT MATRIX AND
COMMUNALITIES OF THE BODY MEASUREMENTS OF ARBOR ACRE, MARSHAL AND ROSS STRAINS OF BROILER CHICKEN.

PC1 PC2 PC3 Communalities

Arbor Acre
Shank length 0.675 0.078 - 0.462
Thigh length 0.789 0.296 - 0.710
Drumstick length -0.245 0.852 - 0.785
Body length 0.141 0.627 - 0.413
Body width 0.105 0.690 - 0.488
Breast width 0.930 -0.130 - 0.883
Wing length 0.897 -0.064 - 0.808
Eigen values 2.841 1.707 -
% of the variance 40.583 24.383 -

Marshal
Shank length 0.885 0.112 0.062 0.800
Thigh length 0.434 -0.736 0.235 0.785
Drumstick length -0.738 0.306 0.106 0.649
Body length -0.028 -0.107 0.939 0.894
Body width 0.634 -0.065 -0.520 0.676
Breast width 0.168 0.899 0.029 0.837
Wing length 0.766 0.010 -0.069 0.592
Eigen values 2.677 1.549 1.007
% of total 38.237 22.132 14.388
variance

Ross
Shank length 0.403 0.824 -0.005 0.842
Thigh length -0.588 0.639 -0.074 0.760
Drumstick length 0.072 -0.244 0.935 0.940
Body length -0.878 -0.009 0.099 0.781
Body width 0.748 -0.144 -0.594 0.933
Breast width 0.913 0.118 -0.247 0.908
Wing length -0.124 0.863 -0.206 0.803
Eigen value 2.733 2,079 1.155
% of total 39.037 29.696 16.501
variance

Two principal components were extracted from Arbor Acre broilers and thigh length (0.824) while PC3 was most highly correlated with
with eigen values of 2.841 for the first principal component (PC1) drumstick length (0.935).
and 1.707 for the second principal component (PC2). The two
principal components accounted for 65% of the total variance DISCUSSION
present in the seven original variables. PC1 had high loadings The average bodyweight of 1.88kg, 1.81kg and 1.65kg attained by
(correlations between the components and the variables) on breast Arbor Acre, Ross and Marshal respectively at 8 weeks of age were
width (0.930), wing length (0.897) and thigh length (0.789). PC2 in line with the report of Akanno et al., (2007) that broiler birds
was orthogonal to PC1 and loaded heavily on drumstick length attain a market weight of 1300.00-2000.00g at 8-10 weeks of age.
(0.852). In Marshal, three principal components were extracted Abdullah et al., (2010) reported an average bodyweight of
with eigen values of 2.677, 1.549 and 1.007 for PC1, PC2 and 1801.00g for Ross broilers at 7 weeks of age. The positive and
PC3 respectively. The three principal components accounted for significant correlations among the body measurements observed
74.76% of the total variance present in the original variables. PC1 in the three strains of broilers indicate high predictability among the
had high positive loadings on shank length (0.885), wing length variables (Pundir et al., 2011).
(0.766) and a negative loading on drumstick length (-0.735)
implying that an increase in shank length and |or wing length will The positive relationship between bodyweight and most of the
results to correlated decrease in drumstick length of Marshal body measurements showed that bodyweight can be predicted
broiler. A similar situation was obtained in PC2 which loaded from body measurements. A similar observation was reported by
heavily on breast width (0.899) and thigh length (-0.736), thus Ajayi et al., (2008). The values of communalities computed for the
implying that an increase in breast width will results to decrease in three strains of chickens confirm that PCA was appropriate for the
thigh length. PC3 was most highly correlated with body length data sets. Yakubu et al., (2009b) reported high range of
(0.939). In Ross broilers, three principal components were communalities (0.755-0.987) for body measurements of Arbor Acre
extracted after varimax rotation accounting for 70% of the total broilers. Mendes (2011) reported a communality range of 0.785-
variance in the original variables with eigen values of 2.733, 2.079 0.987 for body measurements of Ross broilers which was similar
and 1.155 for PC1, PC2 and PC3 respectively. PC1 was most to what was obtained in this study. The lower communality
highly correlated with breast width (0.913), body width (0.748) and observed for body length (0.413) and shank length (0.462) in Arbor
body length (-0.878) implying high relationship among the three Acre imply that the body parameters were weak in explaining the
variables. PC2 had high positive loadings on wing length (0.863) total variation in the body measurements of Arbor Acre broilers.

13

Principal Component Analysis of Body Measurement in Three Strains of Chicken.

Science World Journal Vol 6 (No 2) 2011
www.scienceworldjournal.org
ISSN 1597-6343

Generally in the three strains of chicken, PC1 had the largest Ibe, S. N. (1989). Measure of size and conformation in commercial
share of the total variance and correlated highly with breast width, broilers. Journal of Animal Breeding Genetics. 106: 461-469.
thigh length and shank length. PC1 could be described as the
generalized form of broilers (Salako, 2006). In a principal Jolliffe, I. (2002). Principal Component Analysis. 2nd ed. Springer.
component analysis of body measurements of broilers, Yakubu et
al., (2009b) reported that PC1 had high positive loadings on body Kaiser, H. F. (1960). The application of electronic computers to
weight, breast circumference and thigh length of Arbor Acre and factor analysis. Educational and psychological measurement 20:
termed PC1 “form factor”. Mendes (2011) reported that PC1 had 141-151.
the highest correlation with shank length, breast circumference
and bodyweight of Ross 308 broilers. Yakubu et al., (2009a) Mendes, M. (2009). Multiple linear regression models based on
reported that the first principal component accounted for the principal component scores to predict slaughter weight of broilers.
largest variance in the morphological traits of three Nigerian Archiv Fur Geflugelkunde. 73 (2):139-144.
chicken genotypes. Ogah et al., (2009) presented data that
showed PC1 accounting for the largest variance in the body Mendes, M. (2011). Multivariate multiple regression analysis based
measurements of ducks with high positive loadings on body width, on principal component scores to study relationship between some
bill width, shank length, body length, head length and neck length. pre- and post- slaughter traits of broilers. Journal of Agricultural
Pinto et al., (2006) used PCA to analyze performance and carcass Science (Tarim Bilimleri Dergisi.) 17:77-83.
traits measured in a population of Gallus gallus. The authors
reported that the five first principal components explained 93.30% Monsi, A. (1992). Appraisal of interrelationship among live
of the total variation and the first component explained 66.00%. measurements at different ages in meat-type chickens. Nigerian
They called the first component generalized weight because the Journal of Animal Production 19:15-24.
largest eigen vectors were associated with bodyweight at 35 and
42 days of age, liver, breast, wing and thigh weights. According to Ogah, D. M., Alaga, A. A & Momoh, M. O . (2009). Principal
Mendes (2009), the first principal component provides an component factor analysis of the morph structural traits of
adequate summary of the data in most cases. Muscovy duck. International Journal of Poultry Science 8 (11):
1100-1103.
Conclusion and Recommendation
In the three strains of broiler chicken, PC1 had the largest share of Ogah, D. M. (2011). Assessing size and conformation of the body
the total variance and correlated highly with breast width, wing of Nigerian indigenous turkey. Slovak Journal of Animal Science
length, thigh length and shank length. PC1 could be used to 44 (1): 21-27.
describe the generalized form of broilers. PC2 was orthogonal to
PC1 and loaded heavily on drumstick length, breast width and Pinto, L. F. B., Packer, I. U., De Melo, C. M. R., Ledur, M. C &
wing length. The subsequent component, PC3, was highly Coutinho, L. L. (2006). Principal components analysis applied to
correlated with body length and drumstick length. The three performance and carcass traits in the chicken. Animal Research
principal components could be used to define body size of broilers. 55: 419-425.
These components could be used as selection criteria for
improving meatiness in broilers. The components could also be Pundir, R. K., Singh, P. K & Dangi, P. S. (2011). Factor analysis
used as factor scores for predicting the live weight and carcass of biometric traits of kankrej cows to explain body conformation.
weight of broilers. Asian-Australinian Journal of Animal Science. 24 (4): 449-456.

REFERENCES Rosario, M. F., Silva, M. A. N., Coelho, A. A. D., Savino, V. J. M

Abdullah, Y. A., Al-Beitawi, N. A., Rjoup, M. M. S., Qudsieh, R. I & & Dias, C. T. S. (2008). Canonical discriminant analysis applied to
Ishmais, M. A. A. (2010). Growth performance, carcass and meat broiler chicken performance. Animal 2 (3): 419-424.
quality characteristics of different commercial crosses of broiler
strains of chickens. Journal of Poultry Science. 47: 13-21. Salako, A. E. (2006). Principal component factor analysis of the
morph structure of immature Uda sheep. International Journal of
Ajayi, F. O., Ejiofor, O. & Ironkwe, M. O. (2008). Estimation of Morphology 24 (4): 571-574.
bodyweight from body measurements in two commercial meat-
type chicken. Global Journal of Agricultural Science 7(1): 57-59. SPSS. (2007). Statistical package for the social sciences. SPSS
Inc., 444 Michigan Avenue, Chicago, IL60611, USA.
Akanno, E. C., Ole, P. K., Okoli, I. C. & Ogundu, U. E. (2007).
Performance characteristics and prediction of body weight of Udeh, I.. Ugwu, S. O. C & Ogagifo, N. L. (2011). Predicting
broiler strains using linear body measurements. Proceedings 22nd semen traits of local and exotic cocks using linear body
Annual conference of the Nigerian Society for Animal Production, measurements. Asian Journal of Animal Science 5 (4): 268-276.
Calaber: 162-164.
Yakubu, A., Kuje, D. & Okpeku, M. (2009a). Principal components
Everitt, B. S., Landau, S. & Leese, M. (2001). Cluster Analysis. 4th as measure of size and shape in Nigerian indigenous chickens.
edn. Arnold Publisher, London. Thai Journal of Agricultural Science 42 (3): 167-176.

Eyduran, E., Topal, M & Sonmez, A. Y. (2010). Use of factor Yakubu, A., Idahor, K. O. &. Agade, Y. I. (2009b). Using factor
scores in multiple regression analysis for estimation of body weight scores in multiple linear regression model for predicting the
by several body measurements in brown trouts (Salmo trutta fario). carcass weight of broiler chickens using body measurements.
International Journal of Agriculture & Biology. 12: 611- 615. Revista UDO Agricola 9 (4): 963-967.

14

Principal Component Analysis of Body Measurement in Three Strains of Chicken.