# Determining whether two samples come from the same population and which hypothesis test is appropriate

I have a study conducted by a team of doctors: two groups of people were treated for stress urinary incontinence in two ways: TVT and abbrevo. Each group consisted of 23 people. The hypothesis is as follows:neither method is better than the other.

To begin, I want to check the comparability of groups. That is, to check whether these two samples are a random sample from the same population. Which criteria is best to use?

Secondly, how do I formulate a strictly statistical hypothesis test and what method? Initially I inclined to the Mann-Whitney U test, but Wikipedia says that this criteria can not be used when a lot of duplicates.

1. Age of patient
2. Patients are females
3. Amount of births and pregnancies
4. Mass index
5. Menopause(yes/no)
6. UDI-6 (before operation)
7. IIQ-7 (before operation)
8. Pain (yes/no, before operation)
9. Diagnosis (stress urinary incontinence, recurrent incontinence, or mixed form)
10. Prolapse during op. (yes/no)
11. Complications during operation
12. Postoperative complications
13. Complication (after 6 month)
14. Cough test (after 6 month)
15. Pain (yes/no)
16. UDI-6
17. IIQ-7
-
How was stress urinary incontinence measured? What other data do you have on the people? How were people assigned to TVT vs. abbrevo? –  Peter Flom Oct 28 '12 at 13:21
Patients answered questionaries UID-6 and IIQ-7 and stress urinary incontinence, recurrent incontinence, or mixed forms. Which data do we have is descripted in the question. People were randomly assignated to TVT vs Abbrevo methods –  user31919 Oct 28 '12 at 15:53
If patients were randomly assigned, you don't have to worry too much that they were from the same population. What methods you should use for the main analysis depends on whether UID-6, IIQ-7 and SUI are interval scaled. But it will be some sort of regression. –  Peter Flom Oct 28 '12 at 17:18
I'm sorry i forgot to note that UID-6 and IIQ-7 are not interval scaled but ordinal scaled (patients answer questions how often they have frequent urination from 0 to 3 where 0 means never and 3 means very often) then the answers are sumed into the result of questionnaire. –  user31919 Oct 28 '12 at 17:52
OK, then you should look into ordinal logistic regression. –  Peter Flom Oct 28 '12 at 18:09