DBOLR Test for Testing Autoregressive Conditional Heteroscedasticity and Comparative Study with Two Sided Likelihood Ratio (LR) and Lagrange Multiplier (LM) Tests - A Simulation Approach
Nahida Afroz *
Department of statistics, Comilla University, Comilla, Bangladesh
Hossain Md. Alhelal
Department of Monetary Policy, Bangladesh Bank, Bangladesh
*Author to whom correspondence should be addressed.
Abstract
Many econometric testing problems are potentially either one-sided or partially one-sided because econometric models often come with prior information about the sign of some or all of their unknown parameters. In most cases, time series data suffer from both heteroscedasticity and autocorrelation, which is referred to as autoregressive conditional heteroscedasticity (ARCH) effects. This ARCH usually occurs in financial time series data. Usual two-sided LM and LR type tests are not suitable for testing restricted ARCH effect. In this paper we propose one-sided LR tests for testing ARCH effect in the disturbances of a regression model and compare this with the usual two-sided LR and LM tests. Monte Carlo study indicates that the proposed one-sided LR test performs better than the existing two-sided LR, LM tests.
Keywords: Monte Carlo Simulation, ARCH, LM, LR, Distance-based test