Computer lab 4: Multiple linear regression – further topics: the multicollinearity problem With many possible explanatory variables in a multiple regression model, we often want to choose a model with few variables, if this model is acceptable compared to all other models. In this context we need tests for comparison of models of different size. With many explanatory variables we often have a pronounced problem. The relationship between the response variable and one interesting explanatory variable will be influenced by other explanatory variables, in situations with correlations between the explanatory variables (multicollinearity). The significance and possibility to handle this problem has to be studied. Learning objectives After reading the recommended text and completing the computer lab the student shall be able to: Compare different models with formal tests, Utilize extra sum of squares for these tests; Adopt the concepts of extra sum of squares and partial determination/correlation; Adopt the important concept multicollinearity and make use of methods for discovering this problem. Recommended reading Chapter 7 and Section 10.5 in Kutner et al. Assignment 1: Extra sum of squares and partial F-tests Consider again the data set in exercise 6.18 in the textbook and carry out the exercises 7.7, 7.8 and 7.10. Don’t obtain the p-values of the partial F-tests. Assignment 2: Use the above data and carry out the exercises 7.27 and 10.18. To hand in Answers to assignment 1-2, including computer outputs from 7.27a and 10.18b. The lab report should be handed in no later than 5 days after the scheduled computer lab. Use Lisam (lisam.liu.se) for handing in the assignments.