topic Re: Difference two paired proportions in SAS Programming

Difference two paired proportions

FlorianM — Mon, 05 Mar 2018 12:51:48 GMT

Hello,

I have a binary variable (yes/no) and two measuring time (T1 and T2). Thus data are not independant.

I would like to calculate the difference (in percentage) between the proportion of "yes" at T1 and the proportion of "yes" at T2. Moreover i need to calculate the confidence interval of this difference.

I thought it was possible using proc freq (as paired ttest) but i wasn't successufll at getting it to work.

Hoping that you can help me.

Thank you.

Re: Difference two paired proportions

Ksharp — Mon, 05 Mar 2018 13:08:51 GMT

https://blogs.sas.com/content/iml/2017/07/05/test-equality-two-proportions-sas.html

Re: Difference two paired proportions

FlorianM — Mon, 12 Mar 2018 08:17:38 GMT

Thanks a lot for your response.

Re: Difference two paired proportions

FreelanceReinh — Thu, 07 Feb 2019 11:01:28 GMT

Later readers of this thread should note:

The blog article linked in the accepted solution is about proportions in independent samples. The confidence intervals for a proportion difference in that situation will tend to be too narrow when applied to dependent (paired) samples.

A discussion of CIs for the latter situation (including SAS code) can be found in the SESUG 2013 paper SAS® Macros CORR_P and TANGO: Interval Estimation for the Difference between Correlated Proportions in Dependent Samples.