Selection and validation of parameters in multiple linear and principal component regression for prediction fat-tail weight

The objective of this study was to investigate the relationship among body weight and Fat-tail measurements with fat-tail weight by using Multiple regression and Principal component analysis. The eleven characters includes, body weight, Fat-tail length, Fat-tail circumference, width and diameters in 3 position of upper, middle and lower before the slaughter, and also Fat-tail weight after the slaughter, were measured in 120 Torki-Ghashghaii sheep. After analyzing data, the five first principal components explained 89.69% of the total variability for Fat-tail weight by 45.02, 19.81, 11.58, 7.27 and 6.1, respectively. The highest coefficients in the PC1 and PC2 were to Fat-tail circumference and widths (0.362±0.036) and Fat-tail diameters (0.50±0.34), respectively. Body weight had the highest coefficient in 3 next principal components. Fat-tail length had the highest coefficient in PC3 and PC4. In the Principal component regression, Fat-tail length (0.071), upper circumference of Fat-tail (0.041) and body weight (0.040), had the highest coefficients. While, in General least squares method, middle circumference of Fat-tail (0.083), Fat-tail length (0.077) and body weight (0.042), had the highest coefficients. Also, Principal component regression resulted into much lower amounts for Standard Error (0.006 to 0.02) than General least squares method (0.01 to 0.08). The result of this study showed, estimation of Fat-tail weight using Principal component analysis had a higher accuracy that could be useful for genetic improvement specialists in designing appropriate management, selection and implementation breeding programs.