Is it possible to build a nonlinear mixed-effects model in R with the random effect following a distribution other than Gaussian? For example, a t distribution or a mixed-Gaussian distribution.
To my knowledge, there is no way to directly specify an alternative probability distribution (but maybe others know of something). However, in lme4 the LINK FUNCTION parameters are Gaussian, not the parameters themselves. For example, the random effects in a binomial(logit) model are logit-normally distributed, which looks a lot like a beta distribution when you back-transform them. This makes me wonder if there is a clever way to get, e.g., a mixture of two Gaussian distributions by playing around with the structure of mixed and random effects...