aov and lm

I am trying to fit a model with 4 parameteres on
one response. I have 32 points and the model contains
linear, quadratic and interaction terms. Now when i
apply the Step AIC method most of my square terms drop
out (which i think is not correct if physics is
concerned) So is there any way to force square terms
whicle applying AIC. 

   Also in what way aov() differs anova(lm).

   Is there any other way to select the regressin
model terms?

Thanks in advance.
Vinod


 
____________________________________________________________________________________
Sucker-punch spam with award-winning protection.