Closed romainligneul closed 5 years ago
Hi Romain,
Thanks for spotting this. I will have a look and see with Jean if this is an intended behavior or not.
In the meantime, setting SigmaX0
to zeros(dim.n)
would effectively prevent any update of X0
and properly account for the degrees of freedom.
Hi Romain,
It seems that this is indeed the intended behaviour, as stated in the wiki. Setting the prior variance to 0 is the only effective way to fix the parameters. The various "updatedXX" flags are only internal options to allow partial inference to speed up some iterations and should not be used to specify the model.
Hi Lionel,
The calculation of AIC and BIC values (at least when fitting multinomial data) does not adjust to the option updateX0, so that hidden states are still counted as free parameters even when they cannot be updated.
It is a very minor problem though as they can be recalculated post-hoc using LL.
Romain