Displaying 1 result from an estimated 1 matches for "stepglm3".
Did you mean:
stepglm2
2009 Jan 26
1
glm StepAIC with all interactions and update to remove a term vs. glm specifying all but a few terms and stepAIC
...***
> SNOW:TRI 1.057e-05 3.372e-06 3.135 0.001721 **
The final model included a the TRI:I(TRI^2) term, which is effectively a
cubic function. So this was removed because cubic's were not considered for
all variables. I used update to remove TRI:I(TRI^2). Code:
>stepglm3<-update(stepglm2,~.-TRI:I(TRI^2),trace=T)
> summary(stepglm3)
> extractAIC(stepglm3,k=log(4828))
This results in a 14 term model with a BIC of 3770.172. The BIC is a little
higher, but the cubic term improved fit and is no longer in, so expected.
>Coefficients:
>...