Thanks Joss. Yes I have a feeling the rank approach used in both is behind the same results coming out. But I find it surprising that nowhere is this link mentioned. Hence wondering if SPSS itself may be doing something weird.
------------------------------
David Libbert
------------------------------
Original Message:
Sent: Tue October 31, 2023 06:31 AM
From: Jos Blykers
Subject: Mann Whitney U test - Legacy dialog incorrect SPSS 29.0?
I am not a statistician (but a fan of Andy Field's Discovering Statistics Ed5). Both the Mann-Whitney test and Spearman's rho look at differences in the ranked positions of scores in different groups. Maybe it is not that unusual to find the same results?
------------------------------
Jos Blykers
Original Message:
Sent: Fri October 06, 2023 07:56 AM
From: David Libbert
Subject: Mann Whitney U test - Legacy dialog incorrect SPSS 29.0?
Hello
I've recently found that the Mann Whitney U (MWU) test via the non parametric tests legacy dialogs route can sometimes give different results compared to doing the MHU test via the non parametric tests independent samples route. As well as always having a negative sign for the z score. when in some cases it should be positive (which is correctly reported via the independent samples route), the legacy dialogs route can also come up with a different U statistic compared to the independent samples route.
While I'm on the subject I found it interesting to note that the MWU p values and effect sizes {r=Z/sqrt(n-1)} are exactly the same as that reported by doing a Spearman's correlation (p values and rho). I can't find any reference in the literature linking MWU outputs to Spearman's in this way. Be interested to hear from the community on whether what SPSS is doing is correct/makes sense or is something weird going on?