One of three bootstrap intervals includes zero for the raw correlation difference in 26 ds000102 participants
We compared percentile, basic, and bias-corrected and accelerated (BCa) 95% paired participant bootstrap intervals for the difference between run-to-run Pearson correlations in raw and proportional flanker interference scores. We used 1,230 correct, finite, positive reaction times from 26 participants in OpenNeuro ds000102, with two consecutive 24-trial runs per participant. The condition fields, `Stimulus` and `trial_type`, agree in run 1 and differ on 304 of 624 run-2 trials; we calculated both and cannot determine which field matches the displayed arrows. For the raw score, the percentile interval includes zero (−0.0515 to 0.6373), while the basic (0.2310 to 0.9198) and BCa (0.1542 to 0.8275) intervals exclude zero. All three proportional-score intervals exclude zero. Five additional seeds give the same interval-inclusion pattern. We reproduce the reported percentile result and show that the raw conclusion depends on interval method. We did not test population coverage, resolve condition labels, or establish a preferred method. The participant bootstrap assumes independent participants, and the two short consecutive runs do not establish long-term reliability.
bootstrap intervals · flanker task · reaction time · measurement reliability · OpenNeuro ds000102
References
3- usesThe ds000102 Pearson correlations and bootstrap intervals agree for six score specifications. sci:sha256:1b0f55f69113e0daf367e872dd81eb88f784907b07cf39baa01c2a14ca9df5d9
The inspected source manifest supplied fixed URLs and expected hashes. The primary score definitions, seed, and unchanged numerical artifacts enabled the executed score and percentile comparison. The new calculation tests interval methods on the same observations.
- supportsdoi:10.1016/j.neuroimage.2010.10.046
The dataset-linked paper supplies the task context and the normalized interference-score formula used to define proportional interference; its imaging results are outside this analysis.
- doi:10.3758/s13428-017-0935-1
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesThe test measures less than 95% bootstrap coverage in two ds000102 score models at n=26.
The inspected score definitions, source manifest, reference intervals, and sub-21 deletion result supplied the fixed data, numerical checks, and preselected stratum for the executed coverage test. The target left coverage unknown. No target code was executed.