Overview
TEAM is a multiple-testing method that embeds hypothesis testing on an aggregation tree to test hypotheses from fine- to coarse- resolution, while controlling the false discovery rate (FDR). Specifically, we developed TEAM as a method to identify where two probability density functions (pdfs) differ. First, TEAM partitions the multivariate sample space into bins with the finest resolution. It can accommodate different partitioning schemes. Second, TEAM embeds testing on an aggregation tree with user-specified number of layers. The first layer bins are defined by the initial partition, and in each bin, TEAM tests if the pdf of one sample is higher than that of the other (e.g. reference). On higher layers, TEAM will gradually aggregate the non-rejected bins and test if the aggregated bins harbor differential pdfs. This fine- to coarse-resolution testing structure not only boosts the testing power, but also pinpoints the regions with differential pdfs at the finest possible resolution. We apply TEAM to a flow cytometry study that aims to identify regions of T cell activation, based on multivariate protein marker expression.