Estimation and prediction for a class of dynamic nonlinear statistical models

被引：141

作者：

Ord, JK ^{[1
]}

Koehler, AB

Snyder, RD

机构：

[1] Penn State Univ, Dept Management Sci & Informat Syst, University Pk, PA 16802 USA

[2] Penn State Univ, Dept Stat, University Pk, PA 16802 USA

[3] Miami Univ, Dept Decis Sci & Management Informat Syst, Oxford, OH 45056 USA

[4] Monash Univ, Dept Econometr, Clayton, Vic 3168, Australia

来源：

JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION | 1997年 / 92卷 / 440期

关键词：

forecasting; Holt-Winters method; maximum likelihood estimation; state-space models;

D O I：

10.2307/2965433

中图分类号：

O21 [概率论与数理统计]; C8 [统计学];

学科分类号：

020208 ; 070103 ; 0714 ;

摘要：

A class of nonlinear state-space models, characterized by a single source of randomness, is introduced. A special case, the model underpinning the multiplicative Holt-Winters method of forecasting, is identified. Maximum likelihood estimation based on exponential smoothing instead of a Kalman filter, and with the potential to be applied in contexts involving non-Gaussian disturbances, is considered. A method for computing prediction intervals is proposed and evaluated on both simulated and real data.

引用

页码：1621 / 1629

页数：9