I’m running a hierarchical ordinal regression model, developed along the lines of the hierarchical example in the manual, and the ordered logistic example in the manual.
For some specifications of predictors it was working fine, but for others, I was getting lots of Metropolis rejection messages for the ordered cut-points parameters:
Informational Message: The current Metropolis proposal is about to be rejected because of the following issue:
Exception: Exception: ordered_logistic: Final cut-point is inf, but must be finite! (in 'mot_fits/cat_rat0rs/cat_rat0rs.stan', line 37, column 4 to line 39, column 78) (in 'mot_fits/cat_rat0rs/cat_rat0rs.stan', line 37, column 4 to line 39, column 78)
If this warning occurs sporadically, such as for highly constrained variable types like covariance matrices, then the sampler is fine,
but if this warning occurs often then your model may be either severely ill-conditioned or misspecified.
Applying @betanalpha’s prior on the cutpoints, as he describes in this case study, solved this problem.
I didn’t see this problem documented in the case study or elsewhere, so thought to share.