Negative binomial linear regression models

All the regression models were built using the glm.nb function in R. The final fitted model was determined by performing ANOVA test for models with different independent variables included. We also compared the final negative binomial model with a corresponding Poisson model. P values for the coefficients of included independent variables were calculated using Wald test. To use any model for prediction, 1000 data points for each independent variable were independently simulated. Then the model was used to predict values for the response variable. To remove effects of any independent variable on the response variable, residuals function in R was applied to obtain the corrected values of the response variable.