Plans for multinomial+ BART

[jlevy44]

Are there plans for extending BART responses to binomial, multinomial or ordinal? I would like to make the most use of the conditional mean mu for semiparametric fitting with a mix of the Particle Gibbs and HMC samplers for hierarchical modeling, but can only seem to fit on binomial and continuous outcomes.

Thanks!

[aloctavodia]

Hi, @jlevy44. Yes, there are plans to extend BART to work with those responses. For the moment BART has only been tested for binomial, normal and Poisson/NegativeBinomial responses.

[jlevy44]

Thank you!