A SAS Package for Logistic Two-Phase Studies

Abstract

Two-phase designs, in which for a large study a dichotomous outcome and partial or proxy information on risk factors is available, whereas precise or complete measurements on covariates have been obtained only in a stratified sub-sample, extend the standard case-control design and have been proven useful in practice. The application of two-phase designs, however, seems to be hampered by the lack of appropriate, easy-to-use software. This paper introduces sas-twophase-package, a collection of SAS-macros, to fulfill this task. sas-twophase-package implements weighted likelihood, pseudo likelihood and semi- parametric maximum likelihood estimation via the EM algorithm and via profile likelihood in two-phase settings with dichotomous outcome and a given stratification.