Both intervals included zero for personal start calibration in 26 later YC1 sessions under fixed affine models

We tested whether personal start coefficients from a first session improved spatial response prediction in later YC1 sessions. The source was OpenNeuro ds005522 version 1.0.0. The primary test used 19 people, 26 later sessions, and 1,236 trials. Both affine models could adjust personal target coefficients. Only the full personal model could also adjust start coefficients. With a fixed penalty of 12, mean squared error fell from 200.653 to 196.121 arena units squared. The reduction was 4.532 units, or 2.26 percent. Ten of 19 people had lower error. A bootstrap with 10,000 person draws refitted common maps and kept all sessions for each person together. The 95 percent percentile interval was -0.022 to 11.113 units. The 95 percent basic interval was -2.050 to 9.085 units. Both included zero. The primary result does not establish a reliable benefit or the absence of a benefit. A stronger penalty gave a smaller gain of 0.47 percent in a secondary check. All ten planned secondary checks are reported. The package preserves source extracts, full raw examples, hashes, code, predictions, bootstrap draws, figures, and retrieval instructions. Independent least squares matched the results. All 30 analysis files reproduced exactly. The scope is the available clinical cohort and the fixed models. The analysis does not identify a neural or causal mechanism.

spatial navigation · spatial memory · response prediction · personal calibration · ridge regression · bootstrap uncertainty · OpenNeuro · YC1