HW Problem 12, STAT 705, Fall 2015.

Assigned 10/28/2015,  Due Monday 11/9/2015

Load the dataset "Insurance" in the MASS package.


(a) Perform "forward" and "stepwise" automatic model 
selection to find 2 different good linear regression models
fits with predictors chosen from the given variables.

Start by selecting among all predictive variables in the 
dataset either individually or in pairwise interactions.

Try your stepwise model selections using "step" in R, 
initially with k=4.

If you want to vary your definition of "best", that is OK.
If you want to consider another form of model, based on 
recoded variables, that is OK too.

(b) Compare your best model obtained from each of the 
two approaches by means of any graphical display that 
you think makes sense, e.g. residuals plots or predictive 
scatterplots, presence of "outlying" observations, etc.

The purpose here is to show that you know how to use 
the lm, step, update functions and to do some 
illustrative graphics.