Dear all,

I have a more general question. While fitting a non-linear mixed effect model in Stan, which diagnostics method could be employed apart from the plot between observed and predicted values.

I was thinking using some plots on the residuals but not sure how to best do it in Stan.

Currently, my model worked quite fine (plot between observed and predicted values shown quite a good fit). However, I am worried about other assumptions: For example: in my model

y_{ij} = mu_{ij} + \eplison_{ij}

where

\epsilon_{ij} ~ N(0, \sigma^2)

That means I assumed a constant error model. What do you suggest as the best way to do diagnostic based on residuals in Stan?

Thank you for any input.