Why should we do variable selection? What is the best criterion to use for model selection? Should you use step-wise regression using a criterion or regression subsets with cross-validation? What are the pros and cons of each?

Given the output from lm(), you should be able to say which variable you would want to remove, based on p-value, and also tell us why, in general, this might not be a good approach when you want to reduce the number of variables of the current model?