Normal sleep PVT spread gave an uncertain prediction gain after partial sleep restriction in 45 adults in ds000201 v1.0.3

We tested whether response spread after normal sleep improves a later PVT forecast after assigned partial sleep restriction. PVT means psychomotor vigilance test. We used OpenNeuro ds000201 v1.0.3. We trained two fixed ridge models on 40 adults with the restricted visit first. We tested on 45 other adults whose normal sleep visit came first. The target was the 90th percentile of recorded response times from the first 300 seconds. The base model used normal sleep median response time, age group, and sex. The second model also used log Q90/Q50. Adding this feature reduced mean squared log error by 14.3%. The 95% interval from 20,000 bootstrap draws was -19.5% to 37.2%. One person supplied 88.8% of the net error reduction. This person had spread beyond the training range. The Q80 check reversed the direction. Carrying forward the earlier Q90 gave almost the same log error. These data did not establish an extra prediction benefit. The package preserves source bytes, hashes, the fixed plan, code, model parameters, predictions, bootstrap evidence, figures, and repeat checks. This is a performance forecast test. It does not measure a person's causal response to sleep loss.

psychomotor vigilance test · partial sleep restriction · response time · prediction · sleep deprivation · ridge regression · bootstrap uncertainty