Online Program Home
My Program

Abstract Details

Activity Number: 701
Type: Contributed
Date/Time: Thursday, August 4, 2016 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistics in Marketing
Abstract #318766
Title: Weighted Dirichlet Process Mixture GARCH Model for Predicting Stock Price
Author(s): Inyoung Kim and Peng Sun*
Companies: Virginia Tech and Virginia Tech
Keywords: Back Testing ; Nonparametric Bayesian Model ; Weighted Dirichlet process
Abstract:

In this paper we propose a flexible nonparametric Bayesian approach to analyze Generalized AutoRegressive Conditional Heteroskedasticity (GARCH) model. Our model flexibility is obtained from relaxing the assumption that prior distribution of error term parameters is the same for every observation. Introducing weighted Dirichlet process mixture (WDPM), we can provide multiple candidate priors and let observations to favor different candidates. We model such difference by carefully chosen explanatory covariates and explore different ways of constructing weights. Since GARCH models the log return, it enables us to predict stock price. We develop an empirical way of model evaluation by taking the complexity of real stock trading into consideration and refer it to the Back Testing Return approach. This Back Testing Return approach simulates the real process of stock price prediction and help us to provide decision making. We have conducted large number of model comparisons among parametric and nonparametric Bayesian approaches of stochastic volatility model and nonparametric Bayesian approaches of stochastic volatility model and GARCH using marginal likelihood and using our approach.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

 
 
Copyright © American Statistical Association