Shared habit score covariance exceeded mean crossed baseline covariance in 123 ds004299 adults with an odd and even block split

We compared habit score covariance in 123 adults from OpenNeuro ds004299 version 1.0.0. The scores used extinction cue exposures one and three. The shared calculation used one training baseline in both scores. Crossed calculations used separate odd and even training blocks. Shared covariance was 1.3209. Mean crossed covariance was 1.2767. Their difference was 0.0442, in units of (correct presses per second) squared. The primary 95 percent person bootstrap interval was 0.0311 to 0.0587. The difference was 3.34 percent of shared covariance. Shared Pearson correlation was 0.720. The crossed correlations averaged 0.680. An exact identity gives this covariance difference as one quarter of the baseline difference variance. It holds for finite score vectors without a random error assumption. Across all 200 balanced block partitions, the difference was 3.24 to 10.34 percent of shared covariance. At least 89.66 percent remained in the mean crossed calculation. The crossed scores use fewer blocks per baseline. Their mean correlation is not a corrected reliability estimate. The data do not identify an error fraction or a neural cause of habit. The package preserves 254 source files, fixed source versions, hashes, code, proof, scores, every bootstrap draw, all partitions, and reproduction instructions. Independent calculations and a fresh source replay passed.

habit learning · baseline covariance · measurement · behavioral score · extinction · bootstrap · OpenNeuro