FloSchuberth / cSEM

Composite-based SEM
https://floschuberth.github.io/cSEM/
GNU General Public License v3.0
28 stars 8 forks source link

predict(): Adjust thresholds if two thresholds are exactly the same. #506

Open T-Schamberger opened 1 year ago

T-Schamberger commented 1 year ago

When using OrdPLS(c)predict, it may happen that during cross-validation the data set is split in such a way that two thresholds are exactly equal. This seems to be the case especially with a large number of categories. If two thresholds are exactly equal, this leads to problems when simulating values of the truncated normal distribution for the categorical indicators (line 510). Idea: We can manually adjust the lower or upper threshold value by 0.0001 so that the two values are no longer exactly equal.