What test should I use ? comparing pre/post treatment 3 level categorical data

#1
What test should I use to compare pre/post treatment effect in the form of categorical outcome ?

Post treatment groups have shown decrease in the no. of pads used, what test is most suitable for this scenario? (files attached)

Thanks a lot !
 

Attachments

#2
What test should I use to compare pre/post treatment effect in the form of categorical outcome ?

Post treatment groups have shown decrease in the no. of pads used, what test is most suitable for this scenario? (files attached)

Thanks a lot !
key to know if the data is paired -- I think it is isn 't it ? each individual has a pre and post #pads response.)
If so one option might be the McNemar-Bowker test - if there were more categories (you don't have exact number of the >1 group) then I'd suggest Wilcoxon signed rank but not sure how optimal it is with ties.
(if not paired then Cochran-Armtitage trend test)
tbh the difference seems very clear even without a test for your data - the right test will give pvalue <0.001
 

Karabiner

TS Contributor
#3
What is the topic of your research, and what are the exact research questions?
How was your study designed? Which variables were included and exactly how
were they measured? How large is your sample size?

With kind regards

Karabiner
 
#4
What is the topic of your research, and what are the exact research questions?
How was your study designed? Which variables were included and exactly how
were they measured? How large is your sample size?

With kind regards

Karabiner
My research aims to retrospectively investigate the functional outcome of artificial sphincter on patients with post prostatectomy incontinence (along with complication rates and device survival). The main outcome measured here was in terms of no. of pad use before and after treatment. 45 patients were included in the study.
Thanks for your reply
 
#5
key to know if the data is paired -- I think it is isn 't it ? each individual has a pre and post #pads response.)
If so one option might be the McNemar-Bowker test - if there were more categories (you don't have exact number of the >1 group) then I'd suggest Wilcoxon signed rank but not sure how optimal it is with ties.
(if not paired then Cochran-Armtitage trend test)
tbh the difference seems very clear even without a test for your data - the right test will give pvalue <0.001
yes the data is paired, but some with missing data.
I categorized those with pad use >1 as a single category because those with pad uses >1 were thought of as being still incontinent, where those with 0 pad uses could be thought of as being cured and those with 1 pad use as socially continent.
Glad to take any suggestions. Thanks !
 

Karabiner

TS Contributor
#6
So you have 1 group (n=45) with pre-post data? And your dependent variable is measured as 0, 1, > 1?
You could use the sign test in that case.