UIUCSinhaLab/GEMSTAT

The free_fix (indicator_bool) parser code is broken.

Closed this issue · 6 comments

Investigate, fix.

It looks like the parser code creates an extra entry into indicator_bool that is not present in the input file...
Basically, it doubles the last entry.

Someone should write comprehensive tests to see that it loads the same number of indicator_bools as are needed for the model.
(Different models use different numbers of parameters. :( Thus, really, the par / model should be responsible for this.)

It looks like the factor thresholds don't obey the free_fix file. (And I haven't implemented it for beta yet at all.)

Addendum to last comment : Factor thresholds do obey the free_fix file. Unfortunately ExprPar::adjust does not, so if you specify an out-of-bounds value either in your input par file, the command-line (-et) or the factor_thr_file (-ft), that will get overwritten immediately.

This should be thought of as related to issue #14 .

Fortunately, adjust does not get called if you invoked the program with "-na 0", so this does not affect reproducibility.

Blast. :(

I hate computers.