wwu-mmll / confound_corrected_cpm / 28550949079

85%

main: 85%

Build Type

push

github

Committed by NilsWinter

Commit Message

Fix increment baseline and TFCE threshold rounding; remove dead code

Increment model now computes full - covariates (the connectome's added
value over the confounds), matching the report/docs, instead of
full - connectome. The latter measured the covariates' increment over
the connectome and read ~0 even when the residuals model showed genuine
confound-independent brain signal.

TFCE: add a tolerance to the threshold-sweep comparison so an edge whose
stability lands exactly on a np.arange gridpoint reliably contributes at
its own height. Accumulated arange rounding (e.g. 0.8 -> 0.8000...1) could
otherwise drop an edge's top contribution non-deterministically across
numpy versions, causing pass-locally/fail-on-CI. Adjust the isolated-edge
test to use a sparse H0 null so the property it checks holds under correct
(FP-robust) extent-weighted TFCE.

Remove dead code: calculate_final_cv_results_old (uncallable; referenced
an undefined name), the unused chord_v2.py module (821 lines), and unused
imports.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Coverage Stats

5 of 5 new or added lines in 2 files covered. (100.0%)

55 existing lines in 8 files now uncovered.

2268 of 2713 relevant lines covered (83.6%)

0.84 hits per line