INFORMATION QUANTITIES IN NONCLASSICAL SETTINGS

被引：5

作者：

BARNDORFFNIELSEN, OE ^{[1
]}

SORENSEN, M ^{[1
]}

机构：

[1] AARHUS UNIV,DK-8000 AARHUS,DENMARK

来源：

COMPUTATIONAL STATISTICS & DATA ANALYSIS | 1991年 / 12卷 / 02期

关键词：

ESTIMATING EQUATIONS; INFERENCE FOR STOCHASTIC PROCESSES; MARTINGALES; ROBUSTNESS; SEMIPARAMETRIC MODELS;

D O I：

10.1016/0167-9473(91)90014-S

中图分类号：

TP39 [计算机的应用];

学科分类号：

081203 ; 0835 ;

摘要：

To cope with difficulties met outside the classical framework for likelihood inference various analogue of observed and expected (Fisher) information have been introduced. Here, some of these information quantities are reviewed, exemplified and compared, and their properties are discussed. Besides observed and expected information we consider incremental observed and expected information (in martingale language these are the quadratic variation and quadratic characteristic of the score process), robust observed information, and observed profile information. Extensions of the ideas to estimating equations and to semiparametric models are considered.

引用

页码：143 / 158

页数：16

共 30 条

[1] ON ASYMPTOTIC DISTRIBUTIONS OF ESTIMATES OF PARAMETERS OF STOCHASTIC DIFFERENCE-EQUATIONS [J].