add zero-inflated Poisson/negative binomial distributions

[pkoaz]

I'm interested in using zero-inflated distributions for count data. These are available in brms. I'd be happy to contribute code but I could use some pointers to get started.

[beckyfisher]

Hi Peter, yes zero-inflated models could certainly be useful in some cases. At this stage we don;t have time to implement them, but would be happy to review a PR if you are up to coding it up yourself. Otherwise, if you share an example fitted outside bayesnec using just brms, that would help move things forward when we get a chance to look at it.

In the meantime, you could build the non-linear formula with our package and then run the thing with brms, ie:
fit <- bnec(fvfm ~ crf(log(concentration), model = "nec4param"), data = data)
nl_f <- fit$fit$formula
test <- brms::brm(nl_f, data = data, family = ..., prior = ...)
You would need to construct your own priors but you could also use the priors from the fit as a starting point for that (see our priors vignette, https://open-aims.github.io/bayesnec/articles/example3.html).