Do non-parametric tests generally have more or less statistical power than parametric tests?

Non-parametric tests generally have less statistical power compared to parametric tests when the assumptions of the parametric tests are met. This is because parametric tests usually utilize more information from the data, such as means and variances, allowing them to detect effects more effectively when the data conforms to the assumptions of those tests, like normal distribution.

Non-parametric tests, on the other hand, do not rely on such assumptions and instead use ranks or categories. While they are more robust against violations of assumptions (making them useful in many real-world scenarios), this robustness often comes at the cost of statistical power. As sample sizes increase, the power of non-parametric tests does not increase in the same way that it does for parametric tests if the parametric assumptions are indeed met.

Thus, the general understanding is that while non-parametric tests hold specific advantages under conditions of non-normality or heteroscedasticity, they inherently tend to have less power than parametric tests when the latter’s assumptions are valid.