PREDICTION OF LIVE WEIGHT FROM MORPHOLOGICAL CHARACTERISTICS OF COMMERCIAL GOAT IN PAKISTAN USING FACTOR AND PRINCIPAL COMPONENT SCORES IN MULTIPLE LINEAR REGRESSION
E. Eyduran, A. Waheed*, M. M. Tariq**, F. Iqbal** and S. Ahmad*
Department of Animals Science, Faculty of Agriculture, Igdir University, Turkey; *Bahauddin Zakariya University Multan, Pakistan; **University of Balochistan, Quetta, Pakistan
Corresponding author e-mail: email@example.com or firstname.lastname@example.org
The present work was conducted on 249 local commercial goats in Southern Punjab (Pakistan), particularly Multan, with the aim to define an appropriate model for predicting live weight from linear body measurements recorded from January to June 2012. Data recorded for these goats were body length (BL), chest girth (CG), belly girth BG, rump height (RH), withers height (WH), and live weight (LW). In the statistical evaluation of the data; i) Pearson correlation coefficients between the linear body measurements were estimated, ii) multiple linear regression model was used for live weight prediction, iii) use of the scores from factor and principal component analyses as independent variables in multiple regression analysis model was employed with the purpose to predict live weight (LW). Explanatory variables were body length (BL), chest girth (CG), belly girth (BG), rump height (RH) and wither’s height (WH). The factor scores (FS1, FS2 and FS3) were generated and used as predictors with LW= 0.547 FS1 + 0.313 FS2 – 0.475 FS3. FS1 increased with increasing RH, WH, and CG and indirectly BW would be expected to increase with increasing FS1. When FS2 increased with increasing BG, an increment in BW would be expected. Finally when FS3 increased with increasing BL; an increment in BW would be expected. The obtained equations for principal component analysis were written as: PC1 = – 0.437 BL – 0.462WH – 0.455 RH – 0.423 BG – 0.458 CG, PC2= 0.354 BL + 0.297 WH + 0.261 RH – 0.839 BG – 0.121 CG, LW = - 0.370 PC1 + 0.262 PC2. The original variables largely effective on PC1 were BL, WH, RH, and CG, but BG had a greater effect on PC2 compared with the others. The goat having much greater BL, WH, and RH would be expected to gain greater PC1, PC2 and LW. Goat whose PC1 was negative and absolutely greater value with increasing original variables; viz. BL, WH, RH, BG, and CG would be expected to produce the greater LW. As a result, the present results confirmed that instead of multiple and stepwise regression, principal use of factor and principal component scores in multiple regression analysis might offer a good opportunity without multicollinearity problem for predicting body weight of indigenous goat. Also, if there is a genetic confirmation regarding body weight of commercial goats, their morphological characteristics could provide breeders to acquire some significant clues for further breeding studies.
Key word: Commercial goats, live weight, body measurements, factor analysis.