My understanding is that rstanarm does not yet support any models where the outcome variable is censored. Would it be possible to add this? There are no good frequentist alternatives (eg lme4 does not support censoring and other packages such as lmec are not maintained and very limited in what they can do). Having this resource would be extremely useful for quite a few applications. It may be possible to use the Surv notation to express censoring (left/right/interval).