Binomial regression with stan_mvmer

I’m having trouble fitting a multivariate model using stan_mvmer.
I’m trying to fit two different GLMs, which I assume are correlated. The response variables y_1,y_2 are count vectors, with integer exposures n_1,n_2.
Following this vignette, I’ve defined rates rate_i <- y_i/n_i and want to use the weights parameter to account for the different exposures. I tried:

f <- stan_mvmer(
  formula = list(
    rate_1 ~ feature_2 + feature_3 + (1|feature_1),
    rate_2 ~ feature_2 + feature_4 + (1|feature_1)),
  data = df,
  family = list(binomial, binomial),
  weights = list(df$n_1,df$n_2),
  chains = 1, seed = 12345, iter = 1000)

But when I ran this command, I got nothing; when I commented the weights statement, the model was fitted (but the results were quite meaningless, since the exposures were ignored)

EDIT: I’ve tried using the cbind(y_i,n_i-y_i) syntax and got the following error message:

Error: All outcome values must be 0 or 1 for Bernoulli models.
  • Operating System: Windows
  • rstanarm Version: 2.18.2

It may be that stan_mvmer() doesn’t yet support binomial with trials > 1. @sambrilleman is that right? If so, any plans to allow that?