Livestock Research for Rural Development 22 (9) 2010  Notes to Authors  LRRD Newsletter  Citation of this paper 
2264 data observations on eight body measurements: Head Length, Neck Circumference, Shoulder to Tail Drop, Thoracic Circumference, Body Width, Leg Length, Hip to Knee Length, and Leg Circumference were obtained at birth, weaning (3 weeks), 10, 15 and 20 weeks from 108 local guinea pigs reared on the Teaching and Research Farm of the University of Dschang. Principal component analysis was employed to derive fewer independent common factors in terms of the above mentioned measurements.
Two principal components at birth and one principal component at weaning, 10, 15 and 20 weeks of age accounted for 78%, 76%, 81%, 83% and 95% of the total variance, respectively. The two factors tended to describe the general body size and the appendage of guinea pigs at different ages. The first two extracted factors at birth and the first factor at weaning, 10, 15 and 20 weeks determine the main sources of shared variability that control body conformation in local guinea pigs. These factors could be considered in selection programs to acquire highly coordinated bodies in local guinea pigs with fewer measurements. Principal component analysis is a useful tool in the analysis of the causes of variation observed in the local guinea pig population.
Keywords: body measurements, Cameroon, Guinea pig, PCA, selection, variation
Principal components analysis (PCA) is a multivariate technique for analyzing relationships among several quantitative variables measured on a number of objects, such as persons, soils, fields and plants (Kazutake 1973; Mutsaers et al 1997). It provides information about the relative importance of each variable in characterizing the objects. New variables are calculated which usually consist of linear combinations of the old ones. According to Mutsaers et al (1997), a small number of these new variables will usually be sufficient to describe the observational object without losing too much information. The purpose of factor analysis therefore is to discover simple patterns of relationships among the variables. In particular, it seeks to discover if the observed variables can be explained largely or entirely in terms of a much smaller number of variables which comprise most of the original overall variance called factors.
Takashi and Anthony (1989) employed PCA to extract factors causing soil variation in surface soil samples. They came up with four principal components that accounted for 80.3% of the total variance. Factor analysis with promax rotation for each gender of body measurements in Arabian horses are reported (Sadek et al 2006). This was to derive fewer independent common factors. Three factors were extracted which accounted for 66% and 67% of the total variance in mares and stallions, respectively.
Although principal component analysis (PCA) is a common technique in numerical classification, no researcher has attempted to apply the technique on studies of linear body measurements in guinea pigs. This paper describes the application of PCA to derive fewer independent common factors that could be used to characterize the local guinea pig.
Four boars and 40 does of local guinea pigs were used in the study. They were purchased in local markets in Belo and Bafut in the North West Region, and in Dschang in the Western Region of the Western Highlands of Cameroon. The acquired animals were fairly homogenous in size and shape. They were characterized by tricolored coat pattern; of black and yellow pigmentation with varying degrees of spotting white. The animals were housed in groups of ten does to one boar in suspended bamboo cages measuring approximately 1.2 x 0.7 x 0.6m in a cement block house for seclusion. They were fed ad libitum with fresh forage composed principally of Pennisetum purpureum and Tripsacum laxum, and supplemented daily with a compounded ration. The quantity of supplement given was adjusted to 5% of the average live weight of the animals in each treatment according to Fonteh et al (2005).
Prior to parturition, each doe was transferred to an individual cage measuring 40 x 40 x 60cm to avoid post partum breeding, thereby avoiding inbreeding and to ensure effective monitoring of each sow and its litter. Litters were examined at birth for defects and kids identified with numbered necklace tags and pedigreed with sire and dam. Young animals were nursed by their dams up to weaning (21 days). Weanlings were sexed and separated at 21 days to avoid indiscriminate breeding. This procedure was repeated for the first and second parturitions.
On weaning at 21 days, the does were removed and transferred to the harem while their kids remained in the cages. All cages were cleaned daily to avoid accumulation of urine and faeces. Animals were treated against endoparasites with PIPERAZINE^{®} and with DIDETEKI^{®} against ectoparasites every five weeks, and coccidiosis with an anticoccidian (VETACOX^{®}) every month for three consecutive days according to Fonteh et al (2005). By the time the experiment ended 57 and 35; 48 and 30; 29 and 15; 26 and 13 and 20 and 10 males and females off springs were identified and pedigreed by sire and dam at birth, weaning (3 weeks), 10 weeks, 15 weeks and 20 weeks, respectively.
Linear body measurements (LBMs) that include; Head Length (HL), Neck Circumference (NC), Shoulder to Tail Drop (ST), Thoracic Circumference (TC), Body Width (BW), Leg Length (LL), Hip to Knee Length (HK), and Leg Circumference (LC) were measured in centimeters with a measuring tape as described by Hassan and Chiroma (1991) for linear measurements at the various ages.
The SPSS version 12.0 for word was used for the computation of the principal components. This analyzes the causes of variation of metric traits in guinea pigs, The PCA; a mathematical technique summarizes data and investigates the relationships among variables. For example, a given data set of p variables, some of which may be correlated, PCA transforms them to a new set of p uncorrelated variables called principal components. For each principal component is a linear combination of original variables, whose coefficients are equal to the eigenvectors of the correlation or covariance matrix. The eigenvectors are obtained by the spectral decomposition of the data matrix and arranged in decreasing order of the corresponding eigenvalues, which equal the variances of the components. Thus the first principal component has the largest variance.
This method is generally employed for selecting the first few components which account for most of the variation in the original data as reported by Takashi and Anthony (1989). It helps to understand the data pattern, particularly if some of the original variables are highly correlated. Mathematical operation of PCA has been described by Cooley and Lohnes (1971), and Morrison (1976).
The communalities of variables employed in the PCA are presented in Table 1.
Table 1. Communalities of variables employed in PCA 

Variable 
Birth 
Weaning 
10 Weeks 
15 Weeks 
20 Weeks 
HL 
0.57 
0.54 
0.81 
0.83 
0.99 
NC 
0.73 
0.83 
0.77 
0.64 
0.87 
ST 
0.84 
0.81 
0.89 
0.91 
0.98 
TC 
0.76 
0.74 
0.69 
0.94 
0.99 
BW 
0.77 
0.71 
0.88 
0.93 
0.98 
LL 
0.87 
0.83 
0.85 
0.87 
0.98 
LC 
0.83 
0.85 
0.64 
0.59 
0.94 
HK 
0.79 
0.84 
0.90 
0.91 
0.98 
HL = Head Length; NC = Neck Circumference; ST = Shoulder to Tail Drop; TC = Thoracic Circumference; BW = Body Width; LL = Leg Length, LC = Leg Circumference, HK = Hip to Knee Length 
From this result, relatively high extractions for all the variables studied contributed to factor patterns established.
Tables 2a, 2b, 2c, 2d and 2e show eigenvalues and the proportion of the total variance explained by each of the eight principal components derived from PCA at birth, weaning (21 days), 10, 15 and 20 weeks of age.
Table 2a. Eigen values and the proportion of the total variance for derived principal components at birth 

Principal components 
Eigenvalues 
Proportion 
Cumulative percentage 
PC1 
5.14 
0.64 
64.2 
PC2 
1.01 
0.13 
76.7 
PC3 
0.67 
0.08 
85.1 
PC4 
0.52 
0.07 
91.6 
PC5 
0.29 
0.04 
95.2 
PC6 
0.17 
0.02 
97.3 
PC7 
0.13 
0.02 
99.0 
PC8 
0.07 
0.01 
100.0 
PC = Principal component 
Table 2b. Eigen values and the proportion of the total variance for derived principal components at weaning 

Principal components 
Eigenvalues 
Proportion 
Cumulative percentage 
PC1 
5.95 
0.74 
74.4 
PC2 
0.87 
0.11 
85.3 
PC3 
0.35 
0.04 
90.6 
PC4 
0.25 
0.03 
92.8 
PC5 
0.22 
0.03 
95.5 
PC6 
0.16 
0.02 
97.6 
PC7 
0.11 
0.01 
99.0 
PC8 
0.08 
0.01 
100.0 
PC = Principal component 
Table 2c. Eigen values and the proportion of the total variance for derived principal components at 10 weeks 

Principal components 
Eigenvalues 
Proportion 
Cumulative percentage 
PC1 
6.41 
0.80 
80.2 
PC2 
0.61 
0.08 
87.8 
PC3 
0.35 
0.04 
92.2 
PC4 
0.22 
0.03 
95.0 
PC5 
0.14 
0.02 
96.8 
PC6 
0.12 
0.02 
98.3 
PC7 
0.08 
0.01 
99.3 
PC8 
0.06 
0.01 
100.0 
PC = Principal component 
Table 2d. Eigen values and the proportion of the total variance for derived principal components at 15 weeks 

Principal components 
Eigenvalues 
Proportion 
Cumulative percentage 
PC1 
6.60 
0.83 
82.5 
PC2 
0.69 
0.09 
91.1 
PC3 
0.26 
0.03 
94.3 
PC4 
0.18 
0.02 
96.5 
PC5 
0.13 
0.02 
98.1 
PC6 
0.07 
0.01 
99.0 
PC7 
0.05 
0.01 
99.6 
PC8 
0.03 
0.00 
100.0 
PC = Principal component 
Table 2e. Eigen values and the proportion of the total variance for derived principal components at 20 weeks 

Principal components 
Eigenvalues 
Proportion 
Cumulative percentage 
PC1 
7.70 
0.96 
96.2 
PC2 
0.18 
0.02 
98.2 
PC3 
0.05 
0.01 
98.8 
PC4 
0.04 
0.01 
99.3 
PC5 
0.03 
0.00 
99.6 
PC6 
0.01 
0.00 
100.0 
PC7 
0.00 
0.00 
100.0 
PC8 
0.00 
0.00 
100.0 
PC = Principal component 
The first two principal components (PC1 and PC2) at birth and the first principal component (PC1) at weaning, 10, 15 and 20 weeks accounted for 77%, 74%, 80%, 83% and 96% of the total variance, respectively. The remaining components at both ages became less meaningful and were considered as errors, which included the random components of variation in guinea pigs and various types of errors produced in every stage of sampling and analysis.
Table 3 gives the factor patterns, or loading matrices after varimax rotation and under Kaiser Normalization which characterizes the nature of the first two derived principal components at birth, and the first derived component at weaning, 10, 15 and 20 weeks of age, respectively. Factor patterns consist of the correlation coefficients between the employed variables and the derived principal components.
Table 3. Rotated factor patterns for the first two principal components at birth and the first component at weaning, 10, 15 and 20 weeks of age 

Variable 
Birth 
Weaning 
10 Weeks 
15 Weeks 
20 Weeks 

PC1 
PC2 
PC1 
PC1 
PC1 
PC1 

HL 
0.72 
0.21 
0.74 
0.99 
0.91 
0.99 
NC 
0.85 
0.02 
0.91 
0.88 
0.80 
0.93 
ST 
0.91 
0.08 
0.90 
0.94 
0.95 
0.99 
TC 
0.80 
0.34 
0.06 
0.83 
0.97 
0.99 
BW 
0.86 
0.18 
0.84 
0.94 
0.96 
0.99 
LL 
0.74 
0.57 
0.91 
0.92 
0.93 
0.99 
LC 
0.63 
0.66 
0.81 
0.80 
0.77 
0.97 
HK 
0.86 
0.22 
0.92 
0.95 
0.95 
0.99 
HL = Head Length; NC = Neck Circumference; ST = Shoulder to Tail Drop; TC = Thoracic Circumference; BW = Body Width; LL = Leg Length, LC = Leg Circumference, HK = Hip to Knee Length 
At birth, for the first component, high positive coefficients were given to all the variables studied. These variables correspond to the general body size of the guinea pig. Hence, the first component is considered to determine the general body size of the local guinea pig. The first two derived component at birth are therefore very important in characterizing local guinea pigs without loosing too much details whereas for all the other ages studied, only the first component was essential in effectively characterizing the animal. For the second derived component at birth, positive and negative high coefficients were given to LC and LL, respectively. These variables on their part correspond to appendages of the guinea pig. Apart from TC at weaning that had a low coefficient for the first derived principal component, all other coefficients were high and positive. These could again be describing the general body size of the guinea pig at weaning, 10, 15 and 20 weeks of age. The first two extracted factors at birth and the first extracted factor at any other age considered determine the main sources of shared variability that control body conformation in local guinea pigs. These factors could be considered in selection programs to acquire highly coordinated bodies in the local guinea pig with fewer measurements. It is therefore possible that selection based on superior linear measurements at birth weight and to a lesser extend at weaning can invariably bring about improvements at 10, 15 and 20 weeks of age in guinea pigs in the Western Highlands of Cameroon.
Principal component analysis is a useful tool in the analysis of the causes of variation observed in the local guinea pig population. In our study, variation was decomposed into two factors namely; general body size and appendage factors. The two extracted factors determine the main sources of shared variability that control body conformation in the local guinea pig. These factors could be considered in selection program to acquire highly coordinated bodies in the local guinea pig with fewer measurements. For selection programs intended to bring rapid genetic improvement, focus should be on any of the variables in the first principal component.
Cooley W W and Lohnes P R 1971 Multivariate Data Analysis, 376pp. John Wliey and Sons, New York.
Fonteh F A, Niba A T, Kudi A C, Tchoumboue J and AwahNdukum J 2005 Influence of weaning age on the growth performance and survival of weaned guinea pigs. Livestock Research for Rural Development. Volume 17, Article #133. http://www.lrrd.org/lrrd17/12/font17133.htm
Hassan and Chiroma A 1991 Body weight measurements relationship in Nigerian Red Sokoto goats. Department of Animal Science, Usmanu Danfodio University, Sokoto, Nigeria.
Kazutake K 1973 A method of fertility evaluation for Paddy soils. 111. Third approximation: Synthesis of fertility constituents for soil evaluation.Soil Science and Plant Nutrition 19(1): 1927.
Morrison D F 1976 Multivariate Statistical Methods. 338pp. McGrawHill, New York.
Mutsaers H J W, Weber G K, Walker P and Fischer N M 1997 A Field Guide for OnFarm Experimentation. IITA/CTA/ISNAR http://books.google.fr/books?id=I1aqWjJ5iyoC&printsec=frontcover&dq=A+Field+Guide+for+OnFarm+Experimentation&source=bl&ots=Mw32r0MUo9&sig=c16Z3IZoojmYk2zi8f4JvR5KI&hl=fr&ei=kgsrTPGpFcaisQaN9vXEBA&sa=X&oi=book_result&ct=result&resnum=1&ved=0CBgQ6AEwAA#v=onepage&q&f=false
Sadek M H, AlAboud A Z and Ashmawy A A 2006 Factor analysis of body measurements in Arabian horses. Journal of Animal Breeding and Genetics Vol. 123(6): 369377.
Takashi K and Anthony S R Juo 1989 Multivariate approach to grouping soils in small fields. 1. Extraction of factors causing soil variation by Principal Component Analysis. Soil Science and Plant Nutrition 35(3):469477.
Received 19 April 2010; Accepted 16 June 2010; Published 1 September 2010