Dear all,
I have decided after much deliberation to use backward elimination and
forward selection to produce a multivariate model. Having read about the
problems with choosing selection values I have chosen to base my decisions
of inclusion and exclusion on the AIC and am consequently using the stepAIC
function. This post however does not relate to whether or not this is the
correct decision!
I am interested in determining what the p-value was when a particular
variable was taken out of the model. If I choose trace=TRUE then I
obviously can see each step of the elimination process together with the AIC
and the degrees of freedom for each variable and for the null model. When
the stepwise process is complete it is possible to call the "anova" value
which shows deviances and assocaited degrees of freedoms for variables left
in the model. Therefore I could use this information to calculate
p-values. However, is it possible to do the same for the varaibles which
were thrown out of the model?
There doesn't seem to be any literature on how to use AICs to get p-values
as the distribution isn't quite a chi-squared one. Does anyone therefore
know how I can determine the p-value for a variable when it was taken out of
the model?
Thank you,
Laura
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
