|
Abstract:
|
In a randomized clinical trial, the primary goal is to estimate some measure of treatment effect. Because of limited resources, it is desirable to take advantage of whatever information is available to make as efficient inference as possible. In most large trials, many auxiliary covariates are collected, additional information contained in which may form the basis for more efficient inference. Concerns over the potential for bias are often raised over existing methods for covariate-adjusted analysis. Moreover, such "adjusted" analysis estimate treatment effect conditional on covariates, whereas the primary interest often focuses on the unconditional effect. Using the theory of semiparametrics, we propose a general and flexible approach to exploit auxiliary covariates that leads to more precise estimators for unconditional treatment effect while circumventing the usual concerns.
|
