I'm considering a linear multilevel model. It seems that such a model with two-level coefficient can be estimated by brms. However, I'm wondering if it is possible to allow for a two-level structure in the variance of error term such as $\sigma_i^2 = \sigma^2 + e_i$?
Also, when estimating the same model with brms and Rstan with user-written stan file, which is faster?
