Effect sizes for soscisubmission17 seem off

[LukasWallrich]

@LukasRoeseler in this submission, some correlations are listed as 1 - I noticed that you added a workaround for that in your success criteria code. However, I can generally not quite reproduce the effect sizes, and particularly not the ones listed as 1.

For one, they are not aligned to the correct effects - the infinite odds ratios come from persistent and serious, not from reliable and good-looking. Also, they are infinite because in one condition, 99% rated the person as serious, and in the other 100% - which are infinitely higher odds. Clearly, that is nonsense as that difference is not meaningful - so should we remove those? Or do you know of a better effect size measure we should use instead?

The other odds ratios also do not quite fit (see below)

Here the Table from the original paper

And here ChatGPT's odds ratios:

Trait	Warm	Cold	Odds Ratio
generous	91	8	116.278
wise	65	25	5.571
happy	90	34	17.471
good-natured	94	17	76.490
humorous	77	13	22.405
sociable	91	38	16.497
popular	84	28	13.500
reliable	94	99	0.158
important	88	99	0.074
humane	86	31	13.673
good-looking	77	69	1.504
persistent	100	97	Inf
serious	100	99	Inf
restrained	77	89	0.414
altruistic	69	18	10.140
imaginative	51	19	4.437
strong	98	95	2.579
honest	98	94	3.128

[LukasRoeseler]

For now, I flagged the entries as containing errors and added a "reason for exclusion" to the two infinite effect sizes.