yeatmanlab / jsCAT / 27384496497

100%

main: 100%

Pull #54

github

Document the WLE penalized-form scope trap in the estimator tutorial

The tutorial's penalized objective (max ln L + 0.5 ln I) equals Warm's WLE
only for 1PL/2PL, where Warm's J(theta) coincides with I'(theta). Under 3PL
they diverge (verified against catR source: method = 'WL' solves
S + J/2I = 0 via Ji(), while BM + Jeffreys prior uses I'/2I), producing
theta differences up to ~0.2 on an all-3PL bank and ~0.04 on the golden
fixture bank — beyond the golden test tolerance. The warning spells out the
two valid resolutions so contributors hit this as a documented decision,
not a surprise test failure.

229 of 231 branches covered (99.13%)

Branch coverage included in aggregate %.

100 of 100 new or added lines in 13 files covered. (100.0%)

466 of 466 relevant lines covered (100.0%)

231812.3 hits per line