Pearson-Clopper confidence limits for sensitivity and specific using 2x2 tables in PROC FREQ

[L778967]

Hello,

As the title states, I am trying to calculate Pearson-Clopper confidence limits for sensitivity and specificity using 2x2 crostab tables with PROC FREQ. When I use the SENSPEC option in the TABLES statement, it gives me Wald Confidence limits (https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.5/procstat/procstat_freq_details53.htm). Is there a way I can specify in the SENSPEC option to give me the Pearson-Clopper results? Or does this need to be accomplished another way? Thank you.

[Ksharp]

Pearson-Clopper confidence limits is just EXACT confidence limit.

sensitivity and specific are just performing BINOMIAL distribution test,you can use the following code to get it.

proc freq data=sashelp.heart(obs=1000) ;
table status/binomial(level='Alive');
/* exact binomial;*/
output out=analysis_Sensitivity binomial;
run;

More detail info ,check this NOTE:

https://support.sas.com/kb/24/170.html

[StatDave]

See this note which shows various ways to compute these, and other, statistics and confidence intervals for a 2x2 table. In particular, see the "Other methods to estimate and test the statistics" section. As shown in the first PROC FREQ step for sensitivity, you need to include a WHERE statement to select the level of your response (column) variable that represents the outcome event of interest and then specify your other (row) variable in the TABLES statement with the BINOMIAL option. For specificity, select the nonevent level as shown in the next PROC FREQ step. Use either the EXACT BINOMIAL statement, as shown, or equivalently the CL=EXACT suboption in the BINOMIAL option in the TABLES statement.