3 ms·
It is suspicious but look at the scatterplots: https://www.pnas.org/content/pnas/early/2019/04/24/1901274116.full.pdf#page=4 https://www.pnas.org/content/pnas/e
by gwern 8y ago
It is suspicious but look at the scatterplots: https://www.pnas.org/content/pnas/early/2019/04/24/1901274116.full.pdf#page=4 https://www.pnas.org/content/pnas/early/2019/04/24/190127411... It really is pretty much perfect separation on their metric. When the effect size is that large...
- ufo 8y agoOne thing that would be interesting to investigate in the future would be to see how the test behaves for CFS patients who are not as severely ill, or for non-CFS patients that are currently ill from a different disease.
- mirekrusin 8y agoExacly, just sorting it would show it better, it feels like it should just output percentage with (arbitrary/based on population %age/whatever) threshold that is outside of the test.