# Statistical test choice

#### roentenger

##### New Member
Hi everyone! I'm working on a project and wanted your guy's help since I'm a noob in stats, but excited to learn more.

I'm trying to test the association between two clinical signs in a sample. They all have the disease but they do not all show these signs.

The ideal goal would be to test the diagnostic characteristics of sign A being found (which is dichotomous), but I imagine this is not possible to test since I do not have any healthy controls (all cases have the underlying disease).

Since the former can't be done I would like to describe the characteristics in the small sample and then try to compare it (or find an association) with other dichotomous findings in this sample.

So H0 being, there is no association between sign A and sign (B or C or D or E, all dichotomous). and H1 there being some type of association. What would be the way to test this?

I imagine that testing it in a proportions approach is reasonable. My sample is <30 people. Since this is also dealing with cross findings in a single sample I was thinking that MacNemar test would be the most appropriate. But I am unsure.

Does anyone have any words of advice?

#### Karabiner

##### TS Contributor
McNemar as a test for dependent, binary variables is a reasonable choice.

With kind regards

Karabiner

#### roentenger

##### New Member
Hi Karabiner!
Thanks for your help. If you don't mind me picking your brain further– I tried running the test for sign A and sign B and got the following output.

________________ B
__________ 0 ___1 | Total
--A-------+---------------------
______ 0 | 11____ 0 | 11
______ 1 | 9 ____ 5 | 14
-----------+---------------------
Total___ | 20 ____ 5 | 25
McNemar's chi2(1) = 9.00 Prob > chi2 = 0.0027
Exact McNemar significance probability = 0.0039

Would it be appropriate to say that in these cases finding A's sensitivity was significantly different from that of finding B?