Earlier speed scores gave an uncertain gain for later Digit Symbol prediction in 197 selected DLBS 1.3.0 adults with fixed regression models

We tested whether earlier speed scores improve later Digit Symbol prediction beyond recent scores and fixed background fields. We used 197 selected adults from DLBS 1.3.0. Each test adult was excluded from fitting. Adding two earlier score changes reduced root mean squared error from 6.22 to 6.02 points. Mean squared error gain was 2.49 squared points. The 95% interval from 10,000 refitted person resamples was -1.88 to 8.17. The data allow both a loss and a gain. Seven planned checks also had intervals that included zero gain. One unplanned check removed a partial-task record and gave a similar result. With recent scores present, change terms contain the same information as earlier scores. The test uses selected public records and known visit intervals. It has no external cohort test. The package preserves source files, code, a fixed local plan, input decisions, predictions, all resamples, proofs, and independent checks.

processing speed · Digit Symbol · Digit Comparison · cognitive aging · score prediction · bootstrap · longitudinal data