R vs SAS Confidence Intervals for Paired Proportions

Introduction

See the summary page for general introductory information on confidence intervals for proportions, including the principles underlying the most common methods.

For more technical derivation of methods for comparisons of paired proportions, see the corresponding SAS page.

Since SAS/STAT procedures do not include the direct facility to analyse paired proportions (beyond the McNemar test produced by PROC FREQ), a table of comparison for SAS vs R is omitted here. The Asymptotic Score methods for RD and RR may be obtained in SAS using the %PAIRBINCI macro, which produces results matching the scorepairci() function from {ratesci}. A workaround exists to produce the Wald normal approximation interval for paired RD using pairs as strata, but the performance of this interval is generally poor. Other methods for RD and RR would need to be manually programmed in SAS.

Methods for analysis of conditional (subject-specific) Odds Ratio from paired data are also available in R {ratesci} and {contingencytables}. These could be derived in SAS by transforming a selected method for the single proportion as described here.