Adding earlier lateral ventricle change gave an uncertain 0.86 percent increase in squared forecast error for 88 DLBS 1.3.0 test adults with fixed linear models and a known later scan interval

We tested whether earlier lateral ventricle change improves a forecast of later change in OpenNeuro ds004856 version 1.3.0. A fixed rule selected 176 adults with three MRI scans. We fit two linear models in 88 adults and tested them in 88 other adults. Adding earlier change increased mean squared error from 5.2822 to 5.3275, a 0.86 percent increase. The approximate 95% bootstrap interval for error reduction was -16.51 to 7.65 percent. All seven planned ventricle checks also had intervals that included zero. Both primary models had more test error than a forecast based on the fitting group mean. A paired hippocampal reference reproduced the prior test errors. Separate XML reading and QR calculations passed for all 17,168 numeric fields and 45,000 bootstrap comparisons. Fresh public source files reproduced all six numerical result files exactly. A clean installation also passed the numerical audit and comparison. The package preserves source files, hashes, the local plan, code, forecasts, all sampling draws, figures, and reproduction instructions. The result concerns selected adults, stored MRI measures, fixed models, and a known later scan interval. It does not establish equal performance or the general value of earlier MRI change.

lateral ventricles · longitudinal MRI · forecast evaluation · mean squared error · measurement error · Dallas Lifespan Brain Study