What question did this study set out to answer?

The aim is to evaluate the performance of BIC and UBIC in model selection for autoregressive time series.

March 6, 2026Open Access

Finite sample properties of the Extended Bayesian Information Criterion and related simulations with autocorrected data

Key Points

The aim is to evaluate the performance of BIC and UBIC in model selection for autoregressive time series.
Performed simulations with autoregressive models using FitAR function.
Compared BIC and UBIC for model selection.
Proposed a selection algorithm for regression with autocorrelated errors.
Evaluated performance through subset selection of partial autocorrelations.
Demonstrated consistency in performance of Extended BIC in autoregressive contexts.
Provided insights into effectiveness of model selection algorithms based on BIC and UBIC.

Abstract

Simulations were performed based on McLeod and Zhang's (2008) R package FitAR function SelectModel. Specifically, the BIC (Schwarz, 1978) and the UBIC (Chen Chen, 2008) were compared under time series autoregression models. A model selection algorithm for ordinary regression with autocorrelated error was proposed with both BIC and UBIC. The evaluation of its performance was also conducted. These simulations were performed using a method of subset selection of partial autocorrelations. A short discussion is provided on the consistency property of Chen and Chen's (2008) Extended BIC in the case of autoregression. Finally, an application involving in-hospital respiratory device usage was investigated.

اسأل الذكاء الاصطناعي

Bookmark

View Full Paper