Controlling the Proportion of Falsely Rejected Hypotheses when Conducting Multiple Tests with Climatological Data

• DOI: 10.1175/3199.1
• Publication Date: 2004-11-29T22:50:21Z
• Authors: Valérie Ventura; Christopher J. Pa...; James S. Risbey
• Keywords: False Discovery Rate; Multiple comparisons problem; Independence

Abstract The analysis of climatological data often involves statistical significance testing at many locations. While the field approach determines if a as whole is significant, multiple procedure which particular tests are significant. Many such procedures available, most control, for every test, probability detecting that does not really exist. aim this paper to introduce novel “false discovery rate” approach, controls false rejections in more meaningful way. Specifically, it priori expected proportion falsely rejected out all tests; additionally, test results easily interpretable. also investigates best way apply rate (FDR) spatially correlated data, common climatology. straightforward method controlling FDR makes an assumption independence between tests, while other FDR-controlling methods make less stringent assumptions. In simulation study involving with correlation structure similar real dataset, simple control hypotheses despite violation assumptions, complicated computation little gain alternative hypotheses. A very general no assumptions but cost few Despite its unrealistic assumption, based on results, authors suggest use and provide modification increases power detect