Model comparison using BIC, AIC, -2Log

Hi Luis

I'm not really qualified to comment on this, but I don't think that deciding whether to make an effect fixed or random is a statistical model selection issue. Given that there are issues trying to compare nested models correctly, I can see problems. It's more philosophical, either may be right depending on what you use your model for. Are the levels of A drawn from a larger sample of levels for which you want to make inferences? One issue that may arise is that for factors that are strictly speaking fixed but have alot of levels, it may be easier from a modelling perspective to consider as random. Also you might want to look at some of the papers by Andrew Gelman.

regards

Mike

Luis Orlindo Tedeschi <luis.tedeschi at gmail.com> 04/07/2008 14:23 >>>

Thanks Mike... and I thought it would have a single answer... I glanced
over the link you provided; it will take me some time to digest it. My
current problem is comparing a model with variable A as random effect vs
a model with variable A as fixed effect. It gets vary confusing. Thanks
again. Luis

Dear Luis

It is not necessarily straightforward but there is alot of information out there that can help you. Take a look at http://wiki.r-project.org/rwiki/doku.php?id=guides:lmer-tests and also look through the archives of this list, e.g. the thread entitled "[R-sig-ME] interpreting significance from lmer results for dummies (like me)"

regards

Mike

Luis Orlindo Tedeschi <luis.tedeschi at gmail.com> 03/07/2008 22:23 >>>

Folks; I have a quick question about model comparison. Is it ok to use
BIC/AIC/-2log to compare models with different fixed and random effects
and even different var-(co)var structure? How can I accomplish this
using R? Will Anova do the correct comparison of different models?
Thanks in advance. Luis

-- 
Luis Orlindo Tedeschi <luis.tedeschi at gmail.com>

_______________________________________________
R-sig-mixed-models at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models 

-- 

**+----------------------------------------------------+
              *Luis O. Tedeschi, PhD, PAS
                 Assistant Professor
                 Texas A&M University

230 Kleberg Center               p. (+1) 979-845-5065
2471 TAMU                        f. (+1) 979-845-5292
College Station, TX 77843-2471

*http://nutritionmodels.tamu.edu
http://nutr.tamu.edu 
http://people.tamu.edu/~luis.tedeschi 
+----------------------------------------------------+

This message (and any attachments) is for the recipient ...{{dropped:6}}