A0753
Title: An extension of likelihood ratio based tests for evaluating the interval forecast
Authors: Yushu Li - University of Bergen (Norway) [presenting]
Abstract: Evaluation of the forecast is a fundamental concern in time series forecasting, and the purpose of the evaluation is to check if the ex-post realization aligns with the ex-ante forecast. The likelihood ratio-based tests for evaluating the interval forecast are widely implemented to evaluate the interval forecast, especially in finance risk modelling. For example, value-of-risk (VaR) estimates are an application of one-sided interval forecasting, and the Christoffersen test can be used to test the validity of VaR forecasts in, for example, GARCH models. The likelihood ratio test framework has also been extended to evaluate the density forecast. Although the prior study mentioned their univariate test technique could be extended to multivariate straightforward, no previous study has been done to extend the test to bivariate case time series cases where the forecast area will be a two-dimensional region instead of a one-dimensional interval. The aim is to fill the gap with a concrete extended test framework, investigation of the size and power of the test, and empirical implementations.
