Problems with BIC function about seqhmm

[helske]

This sounds really weird. We do not define BIC function at all, instead we rely on the default BIC method of stats (stats:::BIC.default), which in turn calls the logLikmethod of hmmobject which returns the necessary constants for BIC as attributes. All I can think of is that you have something extra in your environment when running R from shell, like old workspace which defines own BIC.default function, or some other package is loaded with same class name or something?

[vrodriguezf]

Hmmmm I think you are right, probably some of the previous packages I was using to fit HMMs might override the function...Still, I do not understand why it works in RStudio. I will check it and post it here asap.

Thanks a lot!

[vrodriguezf]

Finally, my problem was not exactly on the BIC function, but on the logLik function inside, which I can only link correctly from Rstudio. As a temporal workaround, I have overriden the BIC.default function and make it call explicitly to seqHMM:::logLik

[helske]

I still don't understand how you can have different behaviour with Rstudio and shell if both have same R version at the background with clean workspace... Quick search at R source does not find any bits of code with log(object$K) either...