Questionable research methods: One-sample tests have no power
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
For savings in data collection, Cohen (1977) proposed that instead of a full Neyman-Pearson test (NPTT), only one sample should be collected, and this sample value could be compared to a constant, error-free value. This saves considerable resources around the 75% of the general sample but allows all the information of an NPTT to be used, including the planning of the sample size and the calculation of the power. On closer inspection, this is an epistemological contradiction because you cannot measure error-free. In fact, all these tests are only Fisher tests (NHST). Therefore, nothing can be said about the quality of the experimental condition (α,β,d,n) but only about the probability of the constant´s deviation from the mean of empirical observations under a random distribution (p).