The Department of Statistics and Actuarial Science Fall Colloquium Series presents:

Patrick Breheny, Professor, Department of Biostatistics, College of Public Health, University of Iowa

"A New Perspective on High-Dimensional Intervals"

Abstract: The classical view is that confidence intervals must maintain coverage across all values of the unknown parameter. Inevitably, this breaks down if the underlying estimation procedure is biased. For this reason, many efforts to construct intervals have focused on debiased versions of the lasso. In the process of debiasing, however, the connection to the original lasso estimates is lost. In this work, we offer a different perspective focused on average coverage in contrast to individual coverage. This perspective results in intervals that better reflect the original assumptions that motivate the use of the lasso, as opposed to debiased intervals, which often do not even contain the original lasso estimates. We propose a family of methods that result in approximately correct average coverage along with an efficient computational approach for calculating these intervals and compare these approaches to debiased methods.

Meet and Greet at 3:00 pm in 241 SH. Colloquium at 3:30 pm in 61 SH.