what checks should the cmdstan
diagnose utility impose on a set of inputs?
I ran it on a set of files, carelessly specified as
*.csv, not realizing that there was one file from a different run hanging around - this was the last file to be read in. it had a different value for
max_depth than the others. the stats reported made no sense.
which metadata values need to be the same across input files in order to this utility to work correctly?