Hi,

I am getting this error:

Error in insert_refcat(eta, family = prep$family) :

is.matrix(eta) is not TRUE

When trying to predict **one** new sample, not with more samples, with this model:

```
dat <- data.frame(x=as.factor(c("A", "A", "A", "B", "B", "B")),
y=as.factor(c("A", "A", "B", "B", "B", "A")))
mod1 <- brm(bf(y ~ x), data = dat,
family = categorical(link = "logit"),
iter=3000, warmup=1000, cores=2, chains=2,
control = list(max_treedepth = 11))
newdat <- data.frame(x=c("A","B"), y=c("A", "B"))
posterior_predict(mod1, newdata = newdat, scale="response", nsamples=2) # works
posterior_predict(mod1, newdata = newdat, scale="response", nsamples=1) # gives the above error
```

Perhaps, this is a bug.

- Operating System: Windows 10
- brms Version: 2.13.5
- R version: 3.6.3