Issues in Combining Multiple Genomic Studies

Abstract

As the microarray technology becomes mature and prevalent, increasing number of data sets has been accumulated in the public internet domain. Development for systematic information integration has drawn more attention. In this talk, we will discuss two related methods. We propose an optimally-weighted statistic modified from traditional Fisher's score. We will show its admissibility and its better performance than Fisher's method and minimum p-value method by simulation and applications to real data sets. Currently most methods investigate combining two-class data sets. We will also briefly discuss the issue for combining multi-class studies and propose a solution.