A0736
Title: Prediction for multivariate time series models with deterministic time trends
Authors: YanShuo Pan - National Tsinghua University (Taiwan) [presenting]
Ching-Kang Ing - National Tsing Hua University (Taiwan)
Cy Sin - National Tsing Hua University (Taiwan)
Abstract: When model selection in time series data is referred to, most existing literature considers time series models with a constant mean, while time series data containing deterministic time trends (DTTs) are becoming more common. A multivariate version of the Misspecification-Resistant Information Criterion (MRIC) is proposed for model selection in time series data with DTTS. Unlike the conventional model selection methods, such as Bayesian Information Criterion (BIC) and Akaike Information Criterion (AIC), which focus on correctly specified models, the MRIC proposed by other researchers is designed to cover misspecified models. An asymptotic expansion of the mean squared prediction error (MSPE) is derived in misspecified time series models with DTTs, building on previous work, which provided an asymptotic expression for the MSPE of the least squares predictor. The aim is to show that the multivariate MRIC (MMRIC) achieves asymptotic efficiency regardless of whether the true model is among the candidate models or not. Furthermore, MMRIC can be applied to choose the best multi-step predictive model, which is important for practical applications of time series data.
