ISYE 6414 Final Questions And Answers With Verified Solutions

1. If there are variables that need to be used to control the bias selection in the model, they should forced to be in the model and not being part of the variable selection process. - Answer-True 2. Penalization in linear regression models means penalizing for complex models, that is, models with a large number of predictors. - Answer-True 3. Elastic net regression uses both penalties of the ridge and lasso regression and hence combines the benefits of both. - Answer-True 4. Variable selection can be applied to regression problems when the number of pre- dicting variables is larger than the number of observations. - Answer-True 5. The lasso regression performs well under multicollineariy. - Answer-False 6. The selected variables using best subset regression are the best ones in explaining and predicting the response variables. - Answer-False 8. The lasso regression requires a numerical algorithm to minimize the penalized sum of least squares. - Answer-True 9. An unbiased estimator of the prediction risk is the training risk. - Answer-False 10. Backward and forward stepwise regression will generally provide different sets of selected variables when p, the number of predicting variables, is large. - Answer-True 11. All regularized regression approaches can be used for variable selection. - Answer-False 12. Before performing regularized regression, we need to standardize or rescale the pre- dicting variables. - Answer-True13. The larger the number of predicting variables is, the larger the bias but the smaller the variance is. - Answer-False 14. Variable selection is a simple and solved statistical problem since we can implement it using the R statistical software. - Answer-False 15. BIC penalizes for complexity of the model more than AIC or Mallow's Cp statistics. - Answer-True 16. The penalty constant λ in penalized or regularized regression controls the trade-off between lack of fit and model complexity. - Answer-True 17. We cannot perform variable selection based on the statistical significance of the regression coefficients. - Answer-True 18. Akaike Information Criterion is an estimate for the prediction risk. - Answer-True 19. Forward stepwise regression is a greedy algorithm searching through all possible combinations of the predicting variables. - Answer-False 20. Forward stepwise regression is preferable over backward stepwise regression because it starts with smaller models. - Answer-True 7. The L1 penalty measures the sparsity of a vector. - Answer-True 1. We estimate the regression coefficients in Poisson regression using the maximum likelihood estimation approach. - Answer-True 2. The assumption of constant variance will hold for standard linear regression models with Poisson distributed response data. - Answer-False 3. The F-test can be used to test for the overall regression in Poisson regression. - Answer-False4. The sampling distribution of the prediction of future responses is a t-distribution under the Poisson regression model. - Answer-False 5. We cannot perform goodness-of-fit analysis for logistic regression without replications. - Answer-True 6. We cannot perform a residual analysis for Poisson regression. - Answer-False 7. We can diagnose the constant variance assumption in Poisson regression using the normal probability plot. - Answer-False 8. The estimation of the mean response has higher uncertainty than the prediction of future responses. - Answer-False 9. The expectation of the response variable given the predictors is the sum of the linear combination of the predicting variables in Poisson regression. - Answer-False 10. If a logistic regression provides accurate classification, then we can conclude that it is a good fit for the data. - Answer-False 11. The hypothesis testing procedure for subsets of regression coefficients cannot be used for goodnessof-fit assessment in logistic regression. - Answer-True 12. The logit link function is the only S-shape function that can be used to model binary response data. - Answer-False 13. The standard linear regression model (under the assumption of normality) is not appropriate for modeling binomial response data. - Answer-True 14. The error term in logistic regression has a normal distribution. - Answer-False15. For logistic regression, if the p-value of the deviance test for goodness-of-fit is large, then it is an indication that the model is a good fit. - Answer-True 16. Statistical inference for logistic regression is reliable only for large sample data. - Answer-True 17. The estimated regression coefficients in Poisson regression are approximate. - Answer-True