Exercises for Thursday April 16

1. On April 9 I went through parts of Exam 2009 Exercise 3, in particular the FIC point (c), with 2^2 = 4 submodels, and also most of Exercise 1. Then we discussed the AFIC, where examples & illustrations will be given later. 

2. I have placed "com7b" on the website, which carries out the FIC calculations for the Poisson regression models of Exam 2009 Exercise 3 (c), with the 2^2 = 4 submodels. Please copy this programme to your computer, run it (via "source"), see what happens with different values of (x0, z10, z20), etc.

3. For April 16, construct a suitably modified version of this com7b, for the same dataset of Exam 2009 Exercise 3, but now with 2^3 = 8 candidate models, corresponding to putting extra covariates z1, z2, w in and out, where w is the interaction term w = z1star * z2star, with z1star = (z1 - mean(z1))/sd(z1) and z2star = (z2 - mean(z2))/sd(z2) being the standardised versions of z1 and z2 (thanks to Kathrine FF who pointed us in this direction). Make a FIC table and a FIC plot for the 8 estimates pertaining to (boy, 75, 75). Also compute the AIC value for these 8 models, where you will see, in particular, that it really helps to have interaction on board.

4. Also work through the second part of Exam 2009 Exercise 1, and start with Exercise 2.