I need help in choosing the right test ...

#1
It's been a while since my stats classes ...

I am comparing self-reported data that demonstrates whether or not an intervention was effective at managing symptoms (e.g., yes vs no) from one participant over two unequal periods of time (e.g., opportunities to apply the intervention).
Period 1: yes=32; no=2; total applications=34 (94.12%)
Period 2: yes=22; no=2; total applications=24 (91.67%)

I'm looking for a way to make the case that fewer applications limit the chances of successful management.

What test would you recommend using for this?

Thank you in advance for any help!
 

katxt

Well-Known Member
#4
OK, I see it now, I think.
You could try Fisher's exact test. There are on line calculators. However, I would be very surprised if your data supported the case you hope to prove.
 

katxt

Well-Known Member
#6
You would need a great deal more data (a power analysis would be needed here), and then there you start to get complications like how does one period affect another, so you have problems of lack of independence.
94% and 92% are so close that even if your case could be proved, the clinical importance of that extra 2% is likely so small (I imagine in my ignorance) that your efforts would probably be better spent on some other project with more useful results.