Treatment Effects with Classification Errors - Zi Ye
This event is in the past.
University of Kentucky
Treatment Effects with Classification Errors
In personalized medicine, investigating the differential effect of treatments in groups defined by patient characteristics is of paramount importance. In randomized clinical trials setting, participants are first classified by using diagnostic tools, but such classifiers may not be perfectly accurate. The issue of diagnostic misclassification has been recently become prominent and has been shown to produce severely biased estimations of treatment effect. In this talk, we analyze this problem in a pre-post design. For multivariate continuous outcomes, we propose a multivariate finite mixture to model the misclassification effect. Method for estimating and testing treatment effects, as well as sample size determination, is developed. For ordinal, discrete or skewed outcomes, we develop a fully nonparametric method for estimating and testing treatment effect. Consistent estimators and asymptotic distributions are provided for the misclassification error rates as well as the treatment effect. Simulations study is conducted to compare the new method with traditional methods. The results show significant advantages of the proposed methods in terms of bias reduction, coverage probability and power.