A Monte Carlo Regression Analysis for Interval-Valued Data

• Activity Number: 15
• Type: Topic Contributed
• Date/Time: Sunday, July 29, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Jeongyoun Ahn*+ and Cheolwoo Park and Muliang Peng and Yongho Jeon
• Companies: University of Georgia and University of Georgia and University of Georgia and Yonsei University
• Keywords: Interval-valued data ; Linear regression ; Monte Carlo simulation

Abstract

We consider interval-valued data that frequently appear with the advanced technology in the current data collection processes. Interval-valued data refer to the data observed as ranges instead of single values. In the last decade several approaches to regression analysis have been introduced for interval-valued data, but relevant statistical inferences concerning regression coe?cients have not been studied yet. In this paper, we propose a new approach to ?t a linear regression model to interval-valued data using Monte Carlo sampling. A key advantage is that it enables one to make inferences on the regression coe?cients such as the model signi?cance test, individual slope test and con?dence intervals for slopes. We demonstrate the statistical inference procedure of the proposed approach using simulated and real data, and also compare its performance with those of existing methods.