Testing of assumptions of linear regression model and efficiency of likert type scale to test the assumptions of the model

Linear regression was applied to social science data to study the significant factors contribute to enhancing the farmers' knowledge regarding identifying insect pests and diseases and selecting proper pesticides. Assumptions of the each model were tested to make the model more appropriate for the collected data. To find the Significant Independent Variables, Regression was applied taking farmers knowledge of identifying insect pests, and diseases, and selecting, proper insecticides, pesticides as dependent variable while demographic characteristics, and different sources of information contributing to enhancing the farmers' knowledge as independent variables. Four models were developed each represents the relationship between farmers' knowledge and significant independent variables. To Test the Assumptions of Linear Regression for Likert Type Scale Data. To check the assumptions of linear regression model tested were errors are normally distributed, errors are independently distributed and constant variances of errors. The Residual plot and Non Constant Variance Plot were constructed. Moreover, P-values were also reported for each model to test the null hypothesis. weather there was need of transformation or not. If assumptions were not fulfilled. Therefore, Box and Cox method was applied to get appropriate transformation. The Box and Cox method suggested 0.5, square root transformation. After transformation the model obtained was "QN33=0.397+ 0.016d4 + 0.0007d5 + 0.0188q 332þ. In the model, variable d 4 was not significant because P-value of its slope was greater than 0.05. A regression model was sought having significant independent variables and that fulfills the assumptions of the linear regression. When the model was applied on the actual values, it did not satisfy the assumption of Non-constant Variance. While the model was applied on the transformed values, one independent variables was not significant. To remove this ambiguity, a regression was applied on the actual values of the dependent variable taking two independent variables, ds and 8332 while the third one, d4, was dropped. Stepwise regression suggested the model is given as under: Q33 = 2.093+ 0.006182 d5 + 0.234 q332. The above model reveals that cultivated area under cotton crop, and help from Pesticide Agents had significantly positive effect on the farmers' knowledge regarding selecting the proper insecticides.