I encountered something like this when I used thinning:
Note the selected line, with a relative error count > 1. I sampled for 20,000 samples in 4 chains with thinning=10, thus ending up with 2,000 x 4=8,000 posterior samples. The error count in the selected line shows 11,777 errors: more than were kept.
I think the error summary should respect thinning.
I encountered something like this when I used thinning:
Note the selected line, with a relative error count > 1. I sampled for 20,000 samples in 4 chains with
thinning=10
, thus ending up with 2,000 x 4=8,000 posterior samples. The error count in the selected line shows 11,777 errors: more than were kept.I think the error summary should respect thinning.