Dear List,
I am looking for a function that will find the best subset of negative binomial
models. I have a large data set with 15 variables that I am interested in. I
want an easy way to run all possible models and find a subset of the
"best" models that I can then look at in more detail. I have found
two functions that seem to provide what I am looking for, but am not sure which
one (if either) are appropriate.
glmulti() in package glmulti does an exhaustive search of all models and gives a
number of candidate models to choose from based on your choice of Information
Criterion. This seems to be exactly what I am after, but I found nothing about
it on this list which makes me think there is some reason no one is using it.
gl1ce() in package lasso2 uses the least absolute shrinkage and selection
operator (lasso) to do something. I found it at another thread:
http://tolstoy.newcastle.edu.au/R/help/05/03/0121.html
I did not understand the paper it was based on, and want to know if it even does
what I am interested in before investing a lot of time in trying to understand
it.
Yes, I have read about the problems with stepwise algorithms and am looking for
a valid alternative to narrowing down models when you have a lot of data and a
large number of variables your interested in.
Any thoughts on either of these methods? Or should I be doing something else?
Thanks for your help,
Tim
Tim Clark
Department of Zoology
University of Hawaii