New answers tagged wilcoxon-mann-whitney-test
6
votes
Accepted
Can we interpret the null of the Mann–Whitney U test, i.e. $P(x_i > y_j) = P(y_j > x_i)$, as equivalent to $F = G$ (where $F$ and $G$ are ECDFs)?
If the two distributions are identical, then of course the p-value will be close to 1, as your own answer illustrates.
But the Wilcoxon-Mann-Whitney (WMW) test is only a test on stochastic dominance, ...
- 98k
-2
votes
Can we interpret the null of the Mann–Whitney U test, i.e. $P(x_i > y_j) = P(y_j > x_i)$, as equivalent to $F = G$ (where $F$ and $G$ are ECDFs)?
With Matlab, I tried to verify that, when the null of the Mann-Whitney U-test (MWU) (aka Wilcoxon rank sum (WRS) test), $𝐻_0$: $𝑃(x_i > 𝑦_𝑗) = 𝑃(𝑦_𝑗 > 𝑥_𝑖)$, is satisfied, i.e. when (I ...
- 406
3
votes
Accepted
Null hypotheses of two-sample Kolmogorov-Smirnov, two-sample Anderson-Darling, and (two-sample) Wilcoxon rank sum test
The null for a 2-sample Kolmogorov-Smirnov test (KS) is that the 2 samples come from exactly the same distribution; a KS test of $N(0,1)$ against $N(1,1)$ will be significant (for adequate sample ...
- 3,330
-2
votes
How to choose between t-test or non-parametric test e.g. Wilcoxon in small samples
My answer 10 years after the fact is do the non-parametric equivalent choosing based on sample size and distribution characteristics, don't sweat too much on that but report the effect size! That's so ...
2
votes
Weighted Mann-Whitney U test
This depends on what you mean by "weighted". If you mean sampling weights, then the methods are presented by Lumley & Scott and implemented in R by ...
- 43.5k
Top 50 recent answers are included