2

I have data of 2 groups: 150 vs 70. Can I perform Mann Whitney test as my data was not normally distributed. Should the sample size of two groups be the same in order to run Mann Whitney test ?

Satish Balachanthar
  • 21
  • 1
  • 1
    There's too little information here to judge suitability of Mann-Whitney. Note that it doesn't test the same hypothesis as a t-test (without additional assumptions) but you don't even mention what hypothesis you want to test, nor anything about your variables, nor your sampling. Maybe your variables are 0/1 (in which case I'd suggest something else). Maybe you're comparing a multiset with a subset of it, which would make Mann-Whitney unsuitable as-is. You could be comparing convenience samples. It would be quite irresponsible to jump in with a yes or no on the basis of so little information. – Glen_b Nov 02 '16 at 22:28
  • (The question on whether you need the same sample size can be answered by looking at any set of tables for the test. Googling *Mann-Whitney tables* gets dozens of examples, which all show entries for different sample sizes - like [here](http://ocw.umb.edu/psychology/psych-270/other-materials/RelativeResourceManager.pdf) for example - but I imagine you have some already) – Glen_b Nov 02 '16 at 22:37

1 Answers1

1

Yes. You can use the Mann-Whitney U test for two samples with unequal sample size. However, the statistical power might be limited although it shouldn't matter for you because your sample size is large enough.

Mann-Whitney U test with unequal sample sizes

Community
  • 1
SmallChess
  • 6,764
  • 4
  • 27
  • 48