Activity Number:
|
129
- High-Dimensional Data and Inference
|
Type:
|
Contributed
|
Date/Time:
|
Monday, July 29, 2019 : 8:30 AM to 10:20 AM
|
Sponsor:
|
Biometrics Section
|
Abstract #304427
|
|
Title:
|
A Generalized Framework for High-Dimensional Inference Using Leave-One-Covariate-Out LASSO Path
|
Author(s):
|
Xiangyang Cao* and Karl Gregory and Dewei Wang
|
Companies:
|
University of South Carolina and University of South Carolina and University of South Carolina
|
Keywords:
|
High-dimensional inference;
Variable importance ;
Simultaneous inference;
Bootstrap
|
Abstract:
|
We develop a generalized framework for high-dimensional inference by calculating the change in the LASSO solution path when one covariate is left out. The proposed procedure provides a novel way to calculate variable importance. For high-dimensional linear regression models, our procedure allows for the construction of p-values for testing whether each coefficient is equal to zero as well as for simultaneous testing of multiple hypotheses. Bootstrap techniques are used to construct the null distribution. For low-dimensional linear models, our method has essentially the same power as the t-test. Extensive simulations are provided to show the effectiveness of our method. In the high-dimensional setting, our proposed solution path based test achieves greater power than some other recently developed high-dimensional inference methods. Our method naturally extends to generalized linear models and can be applied to solution paths of other regularization methods.
|
Authors who are presenting talks have a * after their name.