A discrete-continuous mixture quantile function estimator with a practical application to phase II cancer clinical trials

Oftentimes in biological and clinical settings an underlying continuous variable or discrete-continuous mixture is dichotomized in order to interpret the results in a simpler fashion. Tests of interest are then typically based on exact binomial tests. In this article, we illustrate the utility of a new quantile-based approach towards testing the same basic scientific question in terms of improved relative efficiency and similar ease of interpretation when compared with the binomial test. We illustrate the utility of our approach with an eye towards mixtures of discrete-continuous data commonly found in phase II cancer clinical trial data.