Nonparametric test Mann Whitney U

 In Statistics & methods


If you have a numeric variable like salary that you would like to compare with a test between 2 groups say male and females, you might think of using t-test. But there are rules for every test, and here you should have normal distribution in both groups, and in this case it’t NOT normal:

This boxplot above shows outliers both in female and male, and then mean values is not good to use (in t-test mean values are used).

So instead you can use the nonparametric test: Mann Whitney U test:

And in the “Fields” tab you choose the variables to use:

And in the result it’s marked yellow if it’s significant differences between the Groups.

If you double click this result, you will get more information about the test result:

As the significance value (.000) is far below 0.05 we can believe that the salaries differs significant between male and female.

If you want to learn more about which test to choose or how to think when doing analysis, I recommend you to have a look of the trainings we have to offer (see the trainings tab in our portal).

Gunilla Rudander

error: Content is protected !!