|
Activity Number:
|
274
- Macroeconomic Forecasting and Policy in Data Rich Digital Age Environments
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, July 30, 2019 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Business and Economic Statistics Section
|
|
Abstract #300115
|
|
|
Title:
|
Inference in High-Dimensional Models Without Regularization
|
|
Author(s):
|
Ying Zhu* and Kaspar Wuthrich
|
|
Companies:
|
UC San Diego and UC San Diego
|
|
Keywords:
|
nonasymptotic analysis;
high dimensional models;
hypothesis testing;
confidence regions;
limited variability ;
regularization free methods
|
|
Abstract:
|
Regularization methods have become the basis of many inference procedures for high-dimensional models where the number of covariates (p) is larger than or comparable to the sample size (n). I show an empirically relevant setting where regularization based inference methods exhibit substantial omitted variable biases and size distortions. Such a setting concerns limited variability in the covariates, which is common in applied economic research and there are many instances where it occurs “by design”. I then demonstrate situations where regularization-free methods can be adopted or developed for statistical inference in high dimensional regimes. For a new method proposed by Zhu (2019), I also discuss its connection with math programming under uncertainty. This talk is based on the two papers (available on ArXiv): 1. Omitted variable bias of Lasso-based inference methods under limited variability: A finite sample analysis. Kaspar Wüthrich, Ying Zhu (alphabetical ordering. Both authors contributed equally to this work) 2. Statistical inference and feasibility determination: a nonasymptotic approach. Ying Zhu
|
Authors who are presenting talks have a * after their name.
