glum
glum copied to clipboard
Support for quasibinomial, quasipoisson, negative binomial, multinomial & Dirichlet multinomial families?
Are there any plans to also support additional GLM families like quasibinomial, quasipoisson, negative binomial, multinomial, Dirichlet-multinomial (overdispersed multinomial) & ordinal GLMs, some of which are now supported by e.g. h20? (Multinomial & Dirichlet multinomial I believe can be recast into a poisson or quasipoisson GLM via the Poisson trick, but that's computationally not efficient)
No official plans yet, but those are clearly within the scope of this package. The more demand there is for these distributions the more likely they are to be added by the core team or an external contributor.
those are clearly within the scope of this package
The resulting models don't qualify as generalised linear models (because the distributions aren't part of the EDM family), so they aren't clearly within glum's scope, but we could add them nonetheless. :)
I believe we can deal with Negative Binomial first as it is part of EDM family and treat alpha
or theta
parameter of NegBin in the same way like tweedie variance power in current GLUM implementation.