Simultaneous Confidence Intervals for Post Model Selection Inference

• Activity Number: 499 - Nonparametric Multiple Comparison in High Dimensions with Model Uncertainity
• Type: Topic Contributed
• Date/Time: Wednesday, August 2, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Xin Gao*
• Companies: York University
• Keywords: model selection ; inference ; simultaneous interval ; post selection inference ; penalization ; false discovery rate

Abstract

When the model is uncertain, model selection has been often conducted and then based on the selected model, parameters are estimated. Traditional confidence interval for the parameters are based on the assumption that the model being considered is the true model. This approach ignores the uncertainty of the model selection and gives confidence interval which are over optimistic. In this article, we develop the simultaneous confidence intervals for penalized estimates obtained from a penalized estimation strategy.