IMPACT OF INFLUENTIAL AND OUTLIER OBSERVATIONS ON ORDINARY LEAST SQUARE MODEL FOR INDIAN AGRICULTURAL WORKERSBanti Kumar, Manish Sharma*, S. E. H. Rizvi and S.P. Singh
In the present investigation, multiple linear regression model through ordinary least square has been used to study Indian agricultural workers as dependent variable and literacy rate, average size of holding, number of establishments, gross cropped area, net sown area, population density and inflation rate as independent variables. The secondary data related to above variables on all states and union territories have been used for the study purpose. It has been observed that the outliers and influential observations effect the assumptions of ordinary least square. Therefore, the analysis of outliers and influential points becomes an important step of the regression diagnostics which led to different models. Several indicators have been used for identifying outliers and influential observations viz., standardized, studentized and deleted residuals, cooks distance, hat matrix etc. After dropping each influential and outlier observations one by one and in each possible combinations sixty four different models were framed. Best fitted model on basis of minimum mean square error has been proposed and compared with ordinary least square model.
Enter your contact information below to receive full paper.