This document outlines the statistical analysis homework assignment for a food technology course. It involves analyzing two datasets using stepwise model selection with AIC criterion. Students are instructed to: 1) Perform forward selection on the pigs2 dataset considering only main effects, 2) Check assumptions and interpret the final model, 3) Perform backward elimination on pigs2, 4) Explain the differences between R2 and AIC for model selection, 5) Use stepwise selection on the bloodclotting dataset starting with intercept only and main effects only models, 6) Check the final model assumptions, 7) Perform stepwise selection on bloodclotting setting k=0 in the AIC approximation formula.
Original Description:
Exercise for Statistics course. Topic: linear regression, AIC criterion for model selection
This document outlines the statistical analysis homework assignment for a food technology course. It involves analyzing two datasets using stepwise model selection with AIC criterion. Students are instructed to: 1) Perform forward selection on the pigs2 dataset considering only main effects, 2) Check assumptions and interpret the final model, 3) Perform backward elimination on pigs2, 4) Explain the differences between R2 and AIC for model selection, 5) Use stepwise selection on the bloodclotting dataset starting with intercept only and main effects only models, 6) Check the final model assumptions, 7) Perform stepwise selection on bloodclotting setting k=0 in the AIC approximation formula.
This document outlines the statistical analysis homework assignment for a food technology course. It involves analyzing two datasets using stepwise model selection with AIC criterion. Students are instructed to: 1) Perform forward selection on the pigs2 dataset considering only main effects, 2) Check assumptions and interpret the final model, 3) Perform backward elimination on pigs2, 4) Explain the differences between R2 and AIC for model selection, 5) Use stepwise selection on the bloodclotting dataset starting with intercept only and main effects only models, 6) Check the final model assumptions, 7) Perform stepwise selection on bloodclotting setting k=0 in the AIC approximation formula.
Read the help file of the step R function (type ?step).
Download the pigs2 dataset from Minerva (The data are discussed in the slides of Lecture 3). 1. Use the AIC criterion to perform a forward model selection considering only main effects. Discuss the output. 2. Check the assumptions of your final model. 3. Give an interpretation of the model parameters and construct 95 % confidence intervals. 4. Perform now a backward elimination, again using the AIC criterion. The starting model should contain only the main effects. (No need to repeat steps 2 and 3). Download the bloodclotting dataset from Minerva (The data are discussed in the slides of Lecture 3). 5. Explain in a few lines the main difference between using R2 and AIC for selecting models. 6. Use the AIC criterion to perform a stepwise model selection (direction = both, in step R function). Start with a simple model containing only the intercept and and upper model containing only main effects. Discuss the output. 7. Check the assumptions of your final model. 8. Recall that AIC nlog(SSE/n) + kp, for k = 2 (see lecture 3). Setting k = 0, perform a stepwise model selection (direction = both). Start with a simple model containing only the intercept and and upper model containing only main effects. 9. Models obtained in (6) and (9) differ. Briefly explain why. You can put all R output in the appendix. The written report (without the appendix) should not exceed 3 pages.