[Statlist] ETH Young Data Science Researcher Seminar Zurich, Virtual Seminar by Eugene Katsevich, University of Pennsylvania

[Maurer Letizia]

Dear all

We are glad to announce the following talk in the virtual ETH Young Data Science Researcher Seminar Zurich

"Finite-sample optimality and large-sample power analysis of the conditional randomization test"  
by Eugene Katsevich, University of Pennsylvania


Time: Friday, 6 November 2020, 15:00-​16:00
Place: Zoom at https://ethz.zoom.us/j/92367940258

Abstract: For testing conditional independence of a response Y and a predictor X given covariates Z, the recently introduced model-X (MX) framework has been the subject of active methodological research, especially in the context of MX knockoffs and their successful application to genome-wide association studies. In this talk, we study the power of conditional independence testing under MX, focusing our analysis on the conditional randomization test (CRT). The validity of the CRT conditionally on Y,Z allows us to view it as a test of a point null hypothesis involving the conditional distribution of X, from which we can use the Neyman-Pearson lemma to derive the most powerful CRT statistic against a point alternative. We obtain an analogous result for MX knockoffs as well. Next, we derive expressions for the power of the CRT against local semiparametric alternatives, establishing a direct link between the performance of the power of the CRT and the performance of the machine learning method on which it is based. If time permits, we will discuss a recent computational acceleration of the CRT that permits its application to large-scale datasets.

Best wishes,

M. Azadkia, Y. Chen, M. Löffler, A. Taeb

Seminar website: https://math.ethz.ch/sfs/news-and-events/young-data-science.html