Guthery critique of AIC

[sixtystrat]

I know just enough about statistical theory to be dangerous, but would someone care to comment on the invited paper by Guthery et al. in the current issue of JWM? Although I suspect Burnham and Anderson would agree with much of the paper (a lot of criticisms B&A warn us about themselves), I thought some of the comments were a little too "cute".

[Hadu]

Here's another good paper to read on the subject:

Stephens et al. 2005. Information theory and hypothesis testing: a call for pluralism. Journal of Applied Ecology. 42:4-12.

pdf availlable at:

http://www.maths.bris.ac.uk/~mapas/Papers/Stephens%20et%20al%202005%20-%20JApE.pdf


Some folks have problems with the statistical theory underlying AIC, and model averaging based on AIC. See posts at:

http://www.biostat.wustl.edu/archives/html/s-news/2003-07/msg00151.html


djr

[Hadu]

I read the article, thanks for pointing it out. Seems a bit long to make the simple point that your data analyses will not do the thinking for you. Also, no commentary on what I think is a very useful aspect of IC-AIC, model-averaging. Regardless, discussion is good, thoughtful and insightful science is best.