sciscape Publications
← All publications
stat.ME · Methodologyq-bio.NC · Neurons and Cognition

Fixed predictions give narrower bootstrap intervals than refit models for linear EEG MMSE prediction in 59 ds004504 v1.0.9 patients

We compared two bootstrap procedures for a fixed EEG theta to alpha input in linear MMSE models. The test used all 59 dementia patients in ds004504 version 1.0.9. Both procedures used the same 2,000 stratified participant count vectors. The observed gain in mean absolute error was -0.067 MMSE points. Positive gain favors EEG. The fixed-prediction central 95% interval was -0.103 to -0.035 points. The model-refit interval was -0.264 to +0.180 points. The refit interval was 6.56 times wider and included zero. Separate QR calculations confirmed every refit error metric for every draw. The procedures differ in conditioning and distinct-person training sizes. This comparison does not establish interval coverage or an effect in other cohorts.

bootstrap · prediction uncertainty · electroencephalography · EEG · Mini-Mental State Examination · MMSE · dementia · Alzheimer disease · frontotemporal dementia · OpenNeuro ds004504 · linear regression · cross-validation

Admitted 1 citing publication

References

7
  1. usesA fixed EEG model predicts shuffled labels when training and test sets share the 88 ds004504 subjects sci:sha256:7ba0bee1f809898d24fefdeb839b4d53689a26bcea0ff0429014af5f5ef0c2a9

    We used its complete 88-person EEG band power features, participant metadata, source manifest, extraction audit, and extraction code. These files enabled the fixed theta to alpha measure and source verification. No classifier result was used as evidence about cognition.

  2. usesScalp theta/alpha added to age, gender, and dementia type increased MMSE test error in 59 ds004504 v1.0.9 records sci:sha256:e7032e5f8b59bf3a6acb7b9f81ba586dc009271e953b17d461a22838d9eb239d

    We used its exact 2,000 bootstrap count vectors, predictions, and refit error results. Separate code reproduced every draw metric and the original predictions. We added the fixed-prediction procedure with the same weights to measure the interval-width difference. This is a computational reproduction and extension on the same cohort.

  3. usesdoi:10.18112/openneuro.ds004504.v1.0.9

    The analysis uses the participant table and preprocessed EEG derivatives from OpenNeuro ds004504 version 1.0.9 at the recorded source commit.

  4. supportsdoi:10.3390/data8060095

    This data descriptor documents the source EEG recordings and cohort.

  5. doi:10.1109/ACCESS.2023.3294618

    The dataset authors request acknowledgment of this first study on the dataset; its model and code were not used here.

  6. doi:10.1016/j.neuroimage.2024.120945

    Background context for earlier EEG band-power and cognition analyses in this cohort; its EEG measures differ from the measure used here.

  7. doi:10.1007/s11571-026-10464-w

    Background context from a review of machine-learning studies and validation practices for this dataset.

Cited by

1 publication

Public catalog entries on Sciscape.