Statistical power - Parametric vs Nonparametric test

[@Unaimend]

Ok, now I actually watched the whole thing. This was by far the clearest explanation of all those concepts. I would love to take a whole stats course from you. As a computer scientist who got lost in the biology department I have to learn of this by myself and your videos are immensely helpful

[@The111111555555]

Wonderfully explained! For my project, I'm looking for tests that attempted to challenge the statistical power of the good (but ancient) WMW test, similarly when the assumptions are violated and parametric tests can't be used. Would love your guidance of where to look.

[@tilestats]

Permutation tests might be an option: kzbin.info/www/bejne/rGjYaZ9-nNSKn8k

[@MisterDataMoose]

Interesting video, thanks. I tried to run the simulation myself with various levels for the sample size and effect size, and I was quite surprised by how close the power was for the wilcoxon test compared to the t-test. I expected bigger differences. Only for small n and large d, was there a notable difference. On the other hand, when sampling from a lognormal distribution, the wilxocon test clearly had more power in almost all cases. It made me think the wilcoxon test is quite a good default and that there isn't much harm in choosing it even if the data is normally distributed.

[@tilestats]

Yes, the WMW test is more powerful than expected.

[@Unaimend]

Thanks for this video