Mann-Whintey U Test – Statistikproblem?
Abend zusammen,
Ich habe ein Problem mit dem Mann-Whitney U test. Nehmen wir folgende Werte:
A = 8 / 7 / 6 /2 / 5 / 8 / 7 / 3
B = 9 / 9 / 7 / 8 / 10 / 9 / 6
Laut einer Website von der Boston University müssen die U-Werte folgendermaßen sein: U1=46.5 und U2=9.5
Wenn ich die Werte aber in eine Statistik Software eingebe bekomme ich folgende Werte: U1=47.5 und U2=8.5
Weiß jemand warum? Bei anderen Werten funktioniert die Software. Hat die Software irgendwelche Fehler oder mache ich irgendwas falsch. Wie kann ich dieses Problem lösen?
Hier ist der Link zu der Software:
https://www.socscistatistics.com/tests/mannwhitney/default2.aspx
Hier ist der Link zu dem Beispiel und der Lösung der Boston University:
https://sphweb.bumc.bu.edu/otlt/mph-modules/bs/bs704_nonparametric/bs704_nonparametric4.html
(No Ratings Yet)
Loading...
I suspect different procedures with composite values.
“How can I solve this problem?”
What is the problem?
na the different U1 and U2 values, I think
Thanks for the answer. My rehearsal is that the data lists of my investigation are much larger. Consequently, it is not so easy or time-consuming as well as error-prone to manually calculate this test. So I need a reliable software that calculates the results correctly.
I am no longer in the subject, but it could be:
(If several values are identical in both data sets, then theMedianand thearithmetic meansare registered.)
Averaging and median can differ, and you need to know exactly how the median is formed
just seen at socs .
Thanks for the answer. My rehearsal is that the data lists of my investigation are much larger. Consequently, it is not so easy or time-consuming as well as error-prone to manually calculate this test. So I need a reliable software that calculates the results correctly.
You should make a test with a small list you’re getting handy
With 2er and 3er Ties
as in the lower link
Then you can compare it to the top.
The upper seems to be just a money earning link. No real imprint
.
In addition, you should only trust a software that documents the algorithm well.
Even if 2 out of 3 programs show the same, it does not mean that the third is the faulty one.
PS, you already have the almost perfect list 3 times 9 with A and B, 3 times 8
Oh, that’s what you’re asking.
Personally, I would now scrape my documents out of the statistics and quantify it manually, that is not difficult at all. Then you know what values are right.
In general, however, I would type an input error or a wrong setting in the program, so that rankings are misplaced.
Thanks for the answer. My rehearsal is that the data lists of my investigation are much larger. Consequently, it is not so easy or time-consuming as well as error-prone to manually calculate this test. So I need a reliable software that calculates the results correctly.
You could count the test with these values yourself to find out which software was wrong, which hopefully also facilitates troubleshooting.