Skip to content
Prev 31288 / 398506 Next

Stepwise regression?

If you are willing to try lightly tested software, consider "stepAIC.c" 
from "www.prodsyse.com".  This was produced by modifying "stepAIC" from 
MASS.  This can find significant interactions when neither main effect 
is already in the model and forces the linear term in the model when 
testing parabolics.  The output includes an attribute "models" giving a 
posterior distribution over all models tested assuming a uniform prior, 
as described in Burnham and Anderson (2002) Model Selection and 
Multimodel Inference (Springer).

hope this helps.  spencer graves
Petr Pikal wrote: