Model averaging prediction for time series models with a diverging number of parameters

Jun Liao,Guohua Zou,Yan Gao,Xinyu Zhang

DOI: https://doi.org/10.1016/j.jeconom.2020.10.004

IF: 3.363

2021-07-01

Journal of Econometrics

Abstract:<p>An important problem with the model averaging approach is the choice of weights. In this paper, a generalized Mallows model averaging (GMMA) criterion for choosing weights is developed in the context of an infinite order autoregressive (AR(<span class="math"><math>∞</math></span>)) process. The GMMA method adapts to the circumstances in which the dimensions of candidate models can be large and increase with the sample size. The GMMA method is shown to be asymptotically optimal in the sense of achieving the lowest out-of-sample mean squared prediction error (MSPE) for both the independent-realization and the same-realization predictions, which, as a byproduct, solves a conjecture put forward by Hansen (2008) that the well-known Mallows model averaging criterion from Hansen (2007) is asymptotically optimal for predicting the future of a time series. The rate of the GMMA-based weight estimator tending to the optimal weight vector minimizing the independent-realization MSPE is derived as well. Both simulation experiment and real data analysis illustrate the merits of the GMMA method in the prediction of an AR(<span class="math"><math>∞</math></span>) process.</p>

economics,social sciences, mathematical methods,mathematics, interdisciplinary applications

What problem does this paper attempt to address?

This paper attempts to address the following issues:

1. **Weight Selection Problem**: In model averaging methods, how to choose appropriate weights is an important issue. The paper proposes a Generalized Mallows Model Averaging (GMMA) criterion for selecting weights in an infinite-order autoregressive (AR(∞)) process.

2. **Asymptotic Optimality Problem**: It proves that the GMMA method can achieve the lowest out-of-sample Mean Squared Prediction Error (MSPE) in both Independent Realization (IR) prediction and Same Realization (SR) prediction, thereby addressing a conjecture posed by Hansen (2008) on whether the famous Mallows Model Averaging (MMA) criterion is asymptotically optimal when predicting time series.

3. **Multi-step Prediction Problem**: The GMMA method is extended to multi-step prediction, and its asymptotic optimality in multi-step prediction is demonstrated.

4. **Consistency of Weight Estimation**: The paper investigates the rate at which the weight estimator based on the GMMA method converges to the optimal weights that minimize the IR MSPE.

Through theoretical analysis and numerical experiments, the paper demonstrates the advantages of the GMMA method in predicting AR(∞) processes and provides theoretical support for its application.