Adding log mean working memory FD did not establish lower Raven prediction error from PIOP1 to PIOP2 beyond age, sex, and education

We tested a fixed linear prediction procedure in two public AOMIC cohorts. We trained with 199 PIOP1 people and tested with 222 PIOP2 people. The reference model used age, recorded sex, and education. The added model also used log mean working memory FD. FD means framewise displacement. Each motion score used 120 finite values over a fixed 240 s window at 2 s sampling. Adding motion reduced test mean squared error by 0.117 Raven score squared. The 95% interval was [-0.547, 0.685]. The interval used 10,000 independent person bootstrap draws in both cohorts. Both training models were fit again in each draw. All eight planned secondary intervals also included zero. The test did not establish a prediction benefit. It did not prove equal error or absence of motion bias in brain studies. All 431 fresh source motion tables matched the archive. A separate solver checked all primary bootstrap fits. A complete source replay reproduced every result. The package preserves data, code, source records, draws, figures, and reproduction instructions.

framewise displacement · working memory · Raven score · prediction error · AOMIC · cross-cohort prediction