Resting state · dataset and parameters
ABIDE I · NYU controls
This analysis includes 96 NYU control subjects with matched processed runs. It is a defined cohort, not the entire NYU or ABIDE dataset.
The current analysis uses Schaefer 100 cortical ROIs, min–max inputs, equal-subject fitting and separate rank 7, 12 and 20 models. Subject membership and ROI order are identical across these three fits.
Processing and model parameters
| Stage | Current setting |
|---|---|
| Source | ABIDE PCP · C-PAC filt_noglobal functional derivatives |
| Upstream processing | Provider motion correction, nuisance regression, MNI alignment and 0.01–0.1 Hz filtering; no global signal regression in this strategy |
| Local extraction | Timing/geometry checks; nearest-neighbor atlas grid resampling; mask-restricted ROI means; anti-aliased temporal resampling |
| Analysis sampling | 3-second TR; no additional local smoothing |
| Normalization | Per-run, per-ROI min–max to 0–1; historical z-score version retained separately |
| Predictive windows | Past 3 · future 2 |
| Regularization | Ridge 1 |
| Fitted ranks | 7 / 12 / 20, each fitted independently and calibrated for all 96 subjects |
| Reference labels | 7 or 17 Schaefer/Yeo functional systems; all 100 cortical ROIs have assignments |
| Display exclusions | Network 1 hidden as mean-like; the model’s explicit constant bias is also excluded from learned-network displays |
Retained dataset versions
| Atlas | ROIs | Normalization | Runs | Provenance |
|---|---|---|---|---|
| pcp-harvard-oxford-common101 | 101 | minmax | 96 | |
| pcp-harvard-oxford-common101 | 101 | zscore | 96 | |
| schaefer100-cortical | 100 | minmax | 96 |
Quality checks and limitations
11 subjects have parcels below the functional-coverage threshold. The checks retain these warnings; visual alignment and framewise-motion review remain pending.
The cortical atlas excludes subcortex and uses adult-derived parcels across the cohort’s age range. Atlas-label coverage does not validate functional interpretation. These exploratory cohort fits have no held-out prediction, spatial-null significance, or subject-level stability claim.
Saved analysis identity
schaefer100-cortical-minmax--c30dd9dd9577
Checksums and exact settings
Detailed processing record
The scientific preparation is described in Sources, design and methods.More about the measures and interpretation
Grid resampling is not registration. Changing rank does not require extracting ROI time series again.