# Dirichlet prior possible with stanvars?

Hi have a non-linear model that I’d like to fit:

y_{i} \sim \text{Ordered-Logit}(\mu_{i}, \tau) \\ \mu_{i} = \sum_{j = 1}^{J} \beta \phi_{j} x_{t-j} \\ \beta \sim \text{Normal}(0, 0.5) \\ \tau \sim \text{Normal}(0, 1.5) \\ \phi \sim \text{Dirichlet}(1)

Here, all x variables measure the same variable at different points in time. As such, the model treats y as a weighed average of x where \beta serves to convert between units of y and x and the simplex \phi represents the weight placed on each measurement period. For example, if we have 4 periods, so \phi might be {0.7, 0.2, 0.1, 0}.

I have managed to fit the model in Stan, but would like to be able to use the various companion functions that come with brms. I know that dirichlet priors aren’t yet supported in brms. My question: is it possible to fit them anyway using the stanvars() function?

Yes, I think a stanvar for the model block should work. If there is a specific problem you have, feel free to report back. Best of luck!