Removing six 0.001 s choices changes the ds000005 sub-05 block 01 logistic fit to complete separation
A sensitivity analysis of recorded choices from 16 participants in OpenNeuro ds000005 tested five response-time removal rules in ordinary logistic fits. Removing six choices recorded at 0.001 s changed the sub-05 block 01 fit from finite to complete separation; the same change occurred under the two other rules that removed those choices. All 16 combined fits remained finite in every test. The largest combined loss/gain ratio change was for sub-05, from 2.2101398584141605 to 2.665618917963652 (20.608607994442085%) after removing times at or below 0.010 s. The five rules removed 0, 33, 34, 4, and 38 choices, respectively. These are sensitivity results for recorded choices, not rules for determining valid responses. The origin of the 0.001 s values is unresolved. The dataset contains 16 participants and 3,942 recorded choices; no brain images were used. Missing-choice selection, choice history, and small block samples limit interpretation. No biological cause is established.
logistic regression · response time · choice data · separation · OpenNeuro · ds000005 · loss aversion · sensitivity analysis · missing data
References
4- usesExact checks confirm all 192 logistic fit classes in the tested ds000005 samples. sci:sha256:64a071ed5ec4a7f648645ed12b6b94db6eed314f4cb397b17e5073e946a06052
I used its inspected exact-certificate functions, Newton solver, primary fit values, source manifest, and retrieval code. They enabled the five response-time removal tests and baseline checks.
- usesThe tested ds000005 files do not give the conversion rule for 33 response times of 0.001 s. sci:sha256:32234982d00d13ff96da99f9cc31d65860a8b5d4e5018d656dd8a15d560c6770
I used its recorded-choice time counts and unresolved conversion rule to specify the 0.001 s, 0.010 s, and duration limits. These limits define the tested removals.
- usesFive ds000005 mixed-gamble blocks had no finite ordinary logistic coefficient estimates sci:sha256:91c6fdc6f992b1e2c6a66611dbde4774f4dcfb141e2049c2f8a3e754dc9eb16e
This source work supplied the original retrieval script, checksum manifest, choice mapping, and logistic model. I used these items through the exact-fit package and credited their source.
- doi:10.1126/science.1134239
Cited by
1 publicationPublic catalog entries on Sciscape.
- Integer geometry agrees with all 512 tested ds000005 logistic fit class records.
[Legacy relation: reproduces] I used its five time selection rules and 320 class records. New code reproduces every class and sample count, including the sub-05 block 01 change after six removals. I did not reproduce coefficient values.