AIC and AIC distribution discordance

[Ibrahim Giles]

Hello everybody, 

After I computed AIC for my scenarios (M column), I get this results:

M	AIC
Hc9	3669.147
Hc12	3669.835
Hc11	3673.197
Hc10	3677.928

As you can see, the best models were Hc9 and Hc12. Nevertheless, when I used the .par file (with the best parameter values) to estimate AIC distributions for each model, results were a little bit different:

 

While Hc12 showed the AIC distribution with the lowest values, Hc9 was no significantly different from Hc10 (i.e., the worst scenario according to AIC estimation). It strike me that AIC distributions do not support Hc9 as "the second" best model. 

Has anyone else observed these same differences between AIC and AIC distributions?

Cheers

Gustavo Ibrahim Giles P.  

[Hongpu Chen]

Hi there. Have you solved the problem? I get stuck in this situation too.

[Cecilia Fiorini]

Hi!!

Try checking if you a really comparing AIC values... I believe you are plotting likelihoods distributions. As the likelihood does not take into account the number of parameters the results are expected to be different.

Best,

Cecilia

[tavoi...@gmail.com]

Thanks, Cecilia!

Actually I plotted AIC distributions instead of Likelihood. I think this problem could be related with the low number of SNPs that I used. Probably Hongpu has the same issue .

Cheers,

Gustavo I. Giles-Pérez