The Seminar Series Presents:

Max G'Sell, PhD

"Inference after Variable Dependence Structure Estimation"

Abstract: Clustering is often applied to detect dependence structure among the variables in large data sets. However, it is typically difficult to determine the appropriate amount of clustering to carry out in a given instance. In the first part of this talk I will describe a selective inference approach for performing exact tests of dependence structures identifying by hierarchical clustering of variables. This approach yields goodness-of-fit stopping rules for selecting the number of clusters. I will also discuss challenges that arise in extending the inferential procedure through weaker distributional assumptions and more general measures of variable dependence. In the second part of the talk I will demonstrate how similar techniques can be applied to obtain exact tests of graphical models estimated by the graphical lasso. These exact tests provide improved power compared to other popular screening methods for testing dependence between variables, and also give new insight into unexpected behavior of the graphical lasso.

Bio: Max G'Sell is an Assistant Professor of Statistics at in the Carnegie Mellon University Department of Statistics and Data Science. He received his Ph.D. in Statistics from Stanford University in 2014. His interests include statistical methodology and machine learning, especially post-selection inference and clustering, as well as applied work in neuroscience and cellular biology.