Robust model fitting in muzaffarnagri sheep growth data under field conditions

Abstract

Linear regression models were fitted to study the characteristics for growth of Muzaffarnagri sheep breed. Models were fitted with original data separately for male and female animals as well as according to their age. In every case best variables were chosen and used in the models. Data collected under field conditions is liable to have outlying observations. Cook-statistic was applied for detecting the outlying observations. In all the cases 1 or 2 outlying observations were found. Models were fitted again by deleting the outlying observations. Though fitting was improved, yet in some cases it still remained static. This might be due to non-normal distribution of the errors. We therefore, adopted robust regression approach which took care of both outliers and non-normal errors. It was observed that there was drastic improvement in the results, demonstrating the advantage of robust regression fitting in studying the growth features of Muzaffamagri sheep under field conditions.