Seeking methodological review: behavioral instability in ADHD across four open datasets

Announcements
, ,
1

Hello Neurostars community,

I am D1D2DOPAMINE, a pseudonymous independent researcher interested in neuroscience, psychology, open data, and reproducible research.

I am looking for methodological criticism and independent reproduction of an exploratory re-analysis of four public ADHD-related behavioral datasets:

  • HYPERAKTIV
  • BALLADEER
  • UCLA CNP / OpenNeuro ds000030
  • ADHD Pupil Dataset / Figshare 7218725

Background

The project began with a speculative “Allostatic Sprint” hypothesis proposing contrasting Sprint/Crash-like behavioral patterns. As additional datasets were analyzed, the original interpretation was substantially narrowed.

The accumulated evidence does not support stable, task-general Sprint/Crash subtypes, and it does not validate the proposed dopaminergic, metabolic, allostatic, or pupillary mechanism.

The strongest remaining result comes from the ADHD Pupil Dataset. In the primary off-medication ADHD versus control comparison:

  • median reaction time did not differ significantly;
  • robust reaction-time variability (MAD and IQR) was greater in the ADHD group after Holm correction;
  • accuracy was lower after Holm correction;
  • no tested medication endpoint survived multiplicity correction.

I currently interpret this as exploratory, task-dependent evidence for behavioral instability without uniform slowing—not as evidence for discrete subtypes or a biological mechanism.

Feedback requested

I would especially value criticism of:

  1. the permutation-test implementations;
  2. multiplicity correction and endpoint selection;
  3. treatment of missing/no-response trials;
  4. the robustness of the reaction-time variability results;
  5. whether the dimensional behavioral-instability interpretation is justified;
  6. whether the analyses reproduce in a clean environment.

The repository includes the validators, analysis history, null results, corrections, limitations, licenses, and archived outputs.

The work is exploratory, AI-assisted, non-peer-reviewed, and not intended as a diagnostic or clinical tool. I am seeking criticism and reproducibility checks rather than endorsement of the original hypothesis.

Repository:

Archived release:

Independent reproduction and methodological review:

Thank you for taking a look. Even a focused comment on one methodological choice would be useful.