Alternating samples gave smaller cortical connectivity differences than consecutive halves in 20 ds006192 movie runs after filtering of each complete run
We tested whether sample assignment changes cortical connectivity agreement in 20 movie fMRI runs from ten participants. The source was OpenNeuro ds006192 version 1.0.0. Its regional signals had been filtered over each complete run. Both split methods used equal sample counts and the same rows. For 100 cortical regions, the mean RMS difference between Fisher z edge values was 0.00476 for alternating rows and 0.23082 for consecutive halves. The paired difference was 0.22605, with a 95% participant bootstrap interval of 0.21940 to 0.23393. All ten participant contrasts were positive. The difference became much smaller with ten time blocks. Its interval included zero with four blocks. A proof shows equal alternating-part correlation matrices for even periodic signals with Fourier support strictly below one quarter of the sample rate. An independent-channel control met this condition and gave almost exact agreement despite finite-record estimation error. The source Butterworth filter does not meet the strict limit exactly. The proof does not identify the full cause of the observed fMRI difference. Alternating agreement alone cannot establish stability across separate recordings. The package preserves 335 verified source files, code, all planned results, exact split assignments, bootstrap draws, and figures. Offline and fresh-source builds reproduced all 15 numerical result files exactly.
functional connectivity · fMRI · time-series analysis · split-half reliability · temporal filtering · Fourier analysis · OpenNeuro
References
5- usesPosterior 8–13 Hz EEG log ratio agreement was higher for alternating windows than for consecutive halves in 60 adults using 140 planned seconds per eye state sci:sha256:80996e78f08abc36b05ded5322c5525dfd3fbc0fd9e733aaae43983c5597eec5
Its EEG result motivated a comparison of alternating samples and consecutive halves at equal sample counts. This study tests movie fMRI with a different connectivity measure. It reuses no input data or code from that publication.
- usesdoi:10.18112/openneuro.ds006192.v1.0.0
The analysis uses the exact regional signals and metadata from OpenNeuro ds006192 version 1.0.0.
- doi:10.52294/001c.151820
- supportsdoi:10.1002/brb3.456
This study supplies the prior empirical finding cited here that random time-point removal had less effect on functional-connectivity reliability than truncation.
- supportsdoi:10.1016/j.neuroimage.2012.08.022
This study supplies the prior theoretical background cited here on temporal filtering and sample-correlation distributions.
Cited by
7 publicationsPublic catalog entries on Sciscape.
- usesA correction with weight 0.5 reduced mean cortical connectivity prediction error by 6 percent between two movies in ten ds006192 people
We reused 215 selected source files, their retrieval manifest, and the MIT-licensed retrieval algorithm. Its complete-run split report motivated a distinct test of prediction between movies. We recalculated all connectivity vectors and scores, verified public Git hashes, and retrieved every selected source file again.
- usesFourier interpolation gave mean R2 of 0.9934 for alternating cortical samples in 20 ds006192 movie runs processed as complete runs
We used the preserved ds006192 source tables, source manifest, method records, and quality ledger. The earlier connectivity result and strict frequency-limit proof motivated a direct signal recovery test. This package gives new analysis code, target predictions, an error identity without a strict frequency limit, and controls.
- usesFourier phase controls increased alternating correlation differences in 20 movie fMRI runs with fixed channel power spectra
We reused 215 verified source payloads and their provenance. We extended its strict-frequency equality proof to nonzero tails, and reproduced its primary Fisher RMS values in all 20 runs.
- usesAdditive region removal left a positive person match gap between two movies from one visit in ten ds006192 people
We reused its verified ds006192 source archive, atlas table, and fixed source manifest. These files enabled a new person-pairing analysis on the same ten people. We did not reuse its analysis code or claim an independent sample.
- usesThe true pairing had the largest cortical connectivity score across two movies in ten ds006192 participants after movie mean removal
This work used the preserved ds006192 version 1.0.0 source files and retrieval manifest from that publication. Its within-record comparison motivated a separate test of person labels across two movie recordings. The new work tests identity association and does not treat the reused recordings as an independent replication.
- usesPeriodic Fourier interpolation gave a mean squared error ratio of 0.0066 for alternating samples in 100 cortical regions from 20 filtered ds006192 movie runs with odd input lengths
This study used the exact ten-person source subset from this publication. Its alternating connectivity result and strict frequency proof motivated the new sample predictions and finite residual identity.
- usesCortical fMRI patterns matched 8 of 10 people per direction across two movies after movie mean removal in ds006192 v1.0.0
This publication supplied the ds006192 version 1.0.0 source manifest and reported same-person similarity across the two movies. That finding informed the question before the local plan. We downloaded and checked the selected source files again. We tested person links against all other people after movie mean removal. We used the same people, so this study is not an independent replication.