Problems to determine the best model

[aral]

Hello everybody,

I'm using mixed with repeated measurement, and I want to choose the best model, when I compared the different possibilities of the stamen type (ante(1), ar(1), un, cs...) I have some AIC values negatives and positives.

So my question is smaller is better, but smaller means near to 0 or more negative.

One example I have for ante(1) an AIC of -75.8 and for ar(1) 8.8 so which one is the preferred model??

Thank you for the help!

[lvm]

The magnitude of AIC is not meaningful, just the difference in AIC between different models. 0 has no particular meaning. You want the smallest AIC. -75.8 is much smaller than 8.8

[HarryLiu]

It is very important to explore different variance-covariance structures when using proc mixed because the output contains fit statistics indicating which clearly indicate how well each model fits the data compared to other models. You need select the model that has the smallest AIC value.