Need your help with AIC and model selection

fsms

New Member
#1
Hello. If someone could help me with this problem, I would really appreciate it!! I am working on multiple Mantel tests (like multiple regression). As there is no stepwise procedure in most statistical package for this specific test, I decided to choose the best model based on AIC values. However, I didn't get any significant partial correlation (r) if I put more than 2 independent variables, so I decided to compare the significant models with only one variable (Is my idea all right so far?). Then, I don't have to use AIC to choose the best model any more as the number of independent variable is only one? Can I simply choose the best model based on R2 (R square)? :confused:

Thanks very much in advance!
Erika
 
#2
So if you are comparing two matrices with a 3rd, there is no correlation, but if you do one at a time there is?

R^2 is ok but can be a misleading measure, is there any way for you to graph your fitted values vs actuals? What you may also want to look at if possible is something like mean standard error (MSE) to see how much variance your fits generate.