Center and Log Range Models for Interval-Valued Data with an Application to Forecast Stock Returns

• Activity Number: 83
• Type: Contributed
• Date/Time: Sunday, July 31, 2016 : 4:00 PM to 5:50 PM
• Sponsor: International Chinese Statistical Association
• Author(s): Ying Wang* and Yundong Tu
• Companies: Peking University and Peking University
• Keywords: Adaptive estimation ; Forecasting ; Interval-valued data ; Nonlinearity ; Semiparametric models

Abstract

In this paper, we propose to model the interval-valued data using regressions based on center and log range. The advantage of directly modeling log range lies in its simplicity to maintain the natural order of the lower bound and the upper bound of an interval in prediction, without the need to impose any constraint. We further propose two models that incorporate nonlinearities in interval-valued data modeling. In addition, we introduce adaptive estimation to the interval-valued data regressions to deal with the departure from normality. These models are applied to analyze the S&P 500 stock returns, and are found to enjoy superior performance in both the in-sample t and out-of-sample prediction, as compared to the model of Gonzalez-Rivera and Lin (2013).