Relations between information criteria for model-structure selection Part 3. Strong consistency of the predictive least squares criterion
- 1 September 1990
- journal article
- research article
- Published by Taylor & Francis in International Journal of Control
- Vol. 52 (3) , 737-752
- https://doi.org/10.1080/00207179008953564
Abstract
The strong consistency of the predictive least squares (PLS) structure selection criterion is proved for ARMA processes. The advantage of the PLS criterion defined by Rissanen and Wertz (1985) over the BIC, φ or AIC criteria is that it is an online information criterion obtained by simple calculation. The strong consistency is proved for the case of the maximum likelihood estimator, while in simulations the likelihood estimator is substituted by its recursive maximum likelihood estimator approximation. Repeated simulations show the small sample comparison of the PLS and BIC criteria.Keywords
This publication has 4 references indexed in Scilit:
- Recursive Estimation of AutoregressionsJournal of the Royal Statistical Society Series B: Statistical Methodology, 1989
- Universal coding, information, prediction, and estimationIEEE Transactions on Information Theory, 1984
- Rates of convergence for time series regressionAdvances in Applied Probability, 1978
- Estimating the Dimension of a ModelThe Annals of Statistics, 1978