f7ae58c7f1a1cc4abe9273a0f971ba2a-AuthorFeedback.pdf

Thank you for pointing it out. Thatsaid,weunderstand18 that Bernstein-von Mises type results can appear limited when the optimization complications of VB come in. R3 are interested in seeing more details about the simulations, in particular the code, HMC mixing issues, VBoptimizationissues,errorbars,andhighermaximumNvalues. Thank you for the questions. We will include the figure-generation scripts in addition to current Stan code in the28 finalversion.