www.phidot.org

Hello,
Is it legitimate to use AIC as a means of selecting the best model, without addressing model fit using chi-square? My question partly relates to whether these two approaches should be used together (chi-square being a hypothesis-testing approach, and AIC being an information-theoretic approach). Though, I know that chi-square addresses model fit, and AIC addresses model selection, and so they are not used for the exact same process. Still, I was wondering if anyone could comment on the use of AIC in model selection without assessing model fit with chi square. Thanks for any replies!

tpinn wrote:Hello,
Is it legitimate to use AIC as a means of selecting the best model, without addressing model fit using chi-square? My question partly relates to whether these two approaches should be used together (chi-square being a hypothesis-testing approach, and AIC being an information-theoretic approach). Though, I know that chi-square addresses model fit, and AIC addresses model selection, and so they are not used for the exact same process. Still, I was wondering if anyone could comment on the use of AIC in model selection without assessing model fit with chi square. Thanks for any replies!

You might start by reading the first ~5 pages of chapter 5 of the 'MARK book' (in particular, section 5.1).

http://www.phidot.org/software/mark/doc ... /chap5.pdf

Also, the Wikipedia 'article' on AIC is pretty decent -

http://en.wikipedia.org/wiki/Akaike_inf ... _criterion

In particular, see the section on and AIC.