Why isn't there a hurdle gaussian family?

Interfaces brms
1

I’ve been fitting a lot of models on zero-inflated data lately and I’ve been using hurdle models. I find the posterior predictive checks perform better for hurdle models than zero-inflated models for my data in all cases. I’m curious, why isn’t there a family for hurdle gaussian models? Are normally distributed non-zero data just better modeled with a lognormal family? Is there some common wisdom that I don’t know about here?

Thanks all,

K

2

No specific reason. you can build it yourself if you like using custom_family.