search

haziqj / iprobit

Binary and multinomial probit regression using I-priors
GNU General Public License v3.0
1 stars 0 forks source link

Add hyperparameters to the estimation procedure #10

Open haziqj opened 7 years ago

haziqj commented 7 years ago

It might be possible to include things like the FBM Hurst coefficient in the variational Bayes procedure. One caveat might be that the resulting density is not in a recognisable form so maybe something like importance sampling needs to be done.

Incidentally, importance sampling is also needed when dealing with > 3-way interactions or higher order terms.

haziqj commented 6 years ago

This is included already with using a metropolis sampler. Actually, as of today, only the binary sampler works. Still need some work on the multinomial sampler.