Skip to content

AIC and other IT indexes criteria for for backward, forward and stepwise regression

1 message · Szymek Drobniak

#
Hi Mario, firstly - you should always think carefully if model selection is exactly what you want. Secondly - if you have multiple models performing similarly (with deltaAICc <= 2) you can average them (there are several methods, you can e.g. have a look at the MuMIn package) and/or summarize importance of different predictors based on the support they receive in different models.

Cheers,
Szymek

Dr Szymon Drobniak

Institute of Environmental Sciences
Jagiellonian University, Krak?w, Poland

School of Biological, Environmental and Earth Sciences
University of New South Wales, Sydney, Australia

Google Scholar profile
szymekdrobniak.wordpress.com
szdrobniak.pl