4 ms·
Noise. They didn’t have enough data to run the analysis they wanted to run. I would guess they need 50-250 actual hires per interviewer/group to get meaningful
by gleb 6y ago
Noise. They didn’t have enough data to run the analysis they wanted to run. I would guess they need 50-250 actual hires per interviewer/group to get meaningful results.
- nickff 6y agoIt might be noise; I wouldn't try to publish these results in an academic journal, but I'd keep digging. The title uses the word 'superforecasters', which makes me think the author has read some of Philip Tetlock's work, and is trying to apply it to their situation, which I find admirable.
- fractionalhare 6y agoEh, no. I'd expect a dramatic enough effect size that they could make it rigorous with 10 or so actual hires per interviewer. But I agree what's presented here is noise.
- pmiller2 6y agoI agree. You could probably do it with 3-5 hires per interviewer even, if you drastically increased the number of interviewers.