Post #2292064

@olibrendel@scicomm.xyz I would say it's fairly realistic for social sciences, at least. Here effects are often small and measurements noisy. With large survey samples a test may indicate a small effect to be significant at a 5% level, while there may actually be a such effect (a true positive), no real effect (a false positive), or a false positive where the true effect goes in the opposite direction of what's indicated. Related: https://en.wikipedia.org/wiki/Simpson%27s_paradox I was thinking of a situation where test A and test B both test the same hypothesis with e.g. different operationalizations of the outcome variable. In social science a given study often make several tests/measurements of a given phenomenon, and compare the results.