Gustavão, o problema acontece tambem com conjunto de dados grandes! muitas veces se soluciona logaritmizando as variaveis!

Oi Pablo, poxa nos meus modelos só acontece esse erro com N menor que 40. 

Você sabe por que esse erro acontece? 

Abraços,

Gustavo

Just updating.. using the package: 'phylolm' solved the problem. Furthermore, the functions are much faster, which is very important for randomizations. You can check the package here: https://www.google.com.br/webhp?sourceid=chrome-instant&ion=1&espv=2&ie=UTF-8#q=phylolm 

I think this problems surges when the log likelihood  estimate tends to infinity or 0. A way to see if this is happening is explained here: http://www.mpcm-evolution.org/OPM/Chapter5_OPM/OPM_chap5.pdf

essentially you run your PGLS with lambda from 0 to 1 and then graph log likelihood vs lambda. In some cases you can see that as you approach lambda 0 or 1 log likelihood does to infinity.

i will check that package though. 

Hi Cristián,

I checked and this is exactly what is happening.

Thanks for you reply!