Abstract:
|
A common goal of high dimensional genomic data analyses is the development of a class predictor that can be used to assign samples to predefined classes. The class labels may be derived from a binary endpoint or right-censored survival data. Typically in cancer applications, other prognostic markers are available for the samples as well. While fairly standard methods of analyzing such datasets have been developed, sample size methods for such studies are less well established. We present sample size methods we have developed for these settings and discuss various challenges, such as inclusion of clinical covariates into the estimation procedure and computational issues due to the large size and complexity of the calculations.
|