I recall from my stats course that there is a certain n where if it is too small you should just do a non-parametric test regardless but I can't recall what that number was.
Asked
Active
Viewed 26 times
0
-
Related https://stats.stackexchange.com/questions/129755/minimum-sample-size-for-1-way-anova – ralph Feb 11 '21 at 12:04
-
Context would be helpful: what are you trying to achieve? Is this for a homework assignment? A real problem? – Kuku Feb 11 '21 at 18:12
-
1It is characteristic of almost all non-parametric tests that when $n$ is sufficiently small they have no power and are useless. This suggests your recollection might be the opposite of what was taught in the stats course. Regardless, this threshold for $n$ must depend on the test: there is no universal answer. For ANOVA, that means the number of degrees of freedom will be relevant as well as the expected cell means. – whuber Feb 11 '21 at 19:10