Activity Number:
|
338
- Time Series and Forecasting
|
Type:
|
Contributed
|
Date/Time:
|
Tuesday, August 1, 2017 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Business and Economic Statistics Section
|
Abstract #324235
|
View Presentation
|
Title:
|
Time Series Prediction and Predictor Selection in High-Dimensional Quantile Autoregressive Regression
|
Author(s):
|
Dawit Zerom*
|
Companies:
|
California State University at Fullerton
|
Keywords:
|
Conditional Quantile ;
Prediction ;
High Dimensional ;
QAR ;
Penalized ;
FRED database
|
Abstract:
|
With the purpose of multi-step ahead out-of-sample quantile prediction as well as exogenous predictor selection, we introduce a flexible semiparametric quantile regression model that logically extends the quantile autoregressive (QAR) model by allowing arbitrary smooth (possibly nonlinear) functions of high dimensional (possibly larger than the sample size) exogenous predictors. For such partially linear QAR model and assuming exogenous predictor quantile sparsity, we develop a flexible a practical framework to approximate the quantile prediction as well as select relevant exogenous predictors. We also provide an application to U.S. inflation out-of-sample prediction using a set of 134 monthly macroeconomic exogenous predictors variables based on the FRED database.
|
Authors who are presenting talks have a * after their name.