STATS 401 Study Guide - Midterm Guide: Retting, Design Matrix
Document Summary
Due in lab, wednesday, april 21st: the two tables below describe the results of analyzing a ve variable regresison model with sample size n = 50. Y = 1 + x1/2 x3 x4/2 + . The second table gives aic values for all 32 submodels. Solution: the all subsets approach correctly identi es the true model since it has the least aic. The forward selection approach selects the variables in this order: 3,1,4,5,2. The least aic is for the 3,1,4 model which is the population model. The backward selection approach removes the variables in this order: 2,5,4,1,3. The minimum aic is for the 3,1,4 model which is the population model. Solution: variable 4 is almost, but not quite signi cant. So the selected model only includes variables 1 and 3. Therefore this approach does not identify the correct model in this case. 2: suppose the following design matrix and response vector arise in a regression anlaysis:
