The simulation method measures 92.41% to 94.90% coverage for 16 ds000005 ratio models.
In fixed-design simulations based on 16 public ds000005 choice samples, nominal 95% profile-likelihood confidence sets for a loss/gain slope ratio covered the generating ratio in 92.41% to 94.90% of 10,000 simulations per sample. The method retained all 160,000 response samples, including separated samples, by using likelihood suprema. Simultaneous Monte Carlo upper limits were below 95% for four samples. These are conditional simulation results at fitted coefficients. They do not establish coverage in people, across other parameter values, or under dependent choices. The analysis does not calibrate a replacement cutoff or change prior equality-test decisions.
profile likelihood · confidence set coverage · Monte Carlo simulation · logistic regression · separation · loss aversion · fixed design
References
5- usesAll 16 pooled ds000005 ratios had bounded profile intervals, but only 12 excluded one sci:sha256:d5f94cda43c40dba1ba37f6bea8988a179b5f6bfa67c06571c8efd32197d9e1e
Its profile ratio constraint and chi-square-1 cutoff define the confidence-set procedure tested in the fixed-design simulations.
- usesAll binary missing-choice assignments give 10 to 12 Holm p-values less than 0.05 for the ds000005 loss/gain tests. sci:sha256:778e3b5c9d836367300843244c6a2a7187dcea2aeb3ec41ce4a5b7e484d24cd9
The checked source manifest, public text subset, retrieval script, and baseline table enabled source verification and coefficient comparisons.
- usesdoi:10.1126/science.1134239
The publication uses the mixed-gambles task and loss/gain slope-ratio definition described by Tom et al.; it analyzes only the public behavioral text subset.
- doi:10.1037/dec0000141
- usesdoi:10.1214/08-EJS349
The separated-sample calculation uses the likelihood-supremum construction described by Geyer.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.