Mean Difference - Mann Whitney

Discuss statistics related things

by hallgatoemese » Tue Nov 17, 2020 7:42 pm

Hello,

i am wondering how the mean difference is calculated for Mann Whitney U test.
It is not the same as the difference between means, which is okay, since the comparison is not based on means. it is also not the difference between medians (which is also okay, since, strictly speaking, Mann Whitney does not compare medians but mean ranks). However, the number is not even the difference between mean ranks. I have no clue what I am seeing.

Can anyone help me with this? Am I missing something obvious?
Emese
hallgatoemese
 
Posts: 6
Joined: Tue Nov 17, 2020 7:33 pm

by jonathon » Wed Nov 18, 2020 8:38 am

hi emese,

we use the wilcox.test() function from R to calculate this value, but i'm not sure what it represents. here's the R documentation:

https://stat.ethz.ch/R-manual/R-devel/library/stats/html/wilcox.test.html

although it doesn't seem to shed much light either.

jonathon
User avatar
jonathon
 
Posts: 1763
Joined: Fri Jan 27, 2017 10:04 am

by Bobafett » Mon Dec 14, 2020 11:05 am

Are you sure about it NOT being the difference of the medians...?
Image
Bobafett
 
Posts: 18
Joined: Thu Jul 18, 2019 11:33 am

by hallgatoemese » Mon May 10, 2021 8:14 am

Sometimes it does seem like it is the difference of the medians. But it didn't took me long to find an example where it is not the case.

(i attached the screenshot)

Emese
Attachments
mannwhitney-example.png
mannwhitney-example.png (14.49 KiB) Viewed 3359 times
hallgatoemese
 
Posts: 6
Joined: Tue Nov 17, 2020 7:33 pm

by Bobafett » Mon May 10, 2021 1:31 pm

Out of interest - are there any tied ranks in the data? I wonder if they might be affecting these descriptives?
Bobafett
 
Posts: 18
Joined: Thu Jul 18, 2019 11:33 am


Return to Statistics