Description
Hi,
I have a dataset of proportion ranging from 0 to 1, with lots of 0 and 1 data points, which is completely expected from my experiment. From what I've read, it seems that I could use glmmTMB with a zero-one inflated beta to fit the distribution the best (with a mixed model). According to this previous post #355 citing #497 the inflated option seems now possible. But I have an error, even though I'm using the last version of the package (directly downloaded from Github). It is the same error as the one here #355.
Error in eval(family$initialize) : y values must be 0 <= y < 1
Here #497 you wrote something about the problem, but I'm not sure what to do with the "we are defining our own beta_family()
this is a preliminary stab at allowing zero-inflation for otherwise-positive (Gamma, beta) response models. Right now it seems to work OK in a simple beta example. To make it work for Gamma we will have to work around the $initialize() component of the standard Gamma model, which fails unless 0<y<1. (this wasn't too hard for beta because we are defining our own beta_family())
My fit was:
fit_zibeta <- glmmTMB(index_attraction ~ Type * Condition + (1|participant), data=DataCleaned, ziformula= ~1, family=beta_family())
My distribution looks like that:
Any idea what I could do wrong?
Many thanks!