HW 2

Comments
Just a few comments:
  1. When you are trying to find best model, you can not just remove all interaction at once, you have to remove one at a time.
  2. Dust have three levels, so if you are checking difference between two levels at a time, you have to check THREE possible combinations.
  3. If you are specifically asked to check if numeric representation of emp could be used, than it is exactly what you had to do, and not just give a comment, that "we could do it".
  4. If you are using glm functions in Splus, you should not worry about Y been 0 at some point, Splus takes care about it.
Other than that, no problems.