Efficient and adaptive post-model-selection estimators

被引：12

作者：

Bühlmann, P ^{[1
]}

机构：

[1] ETH Zurich, Seminar Stat, CH-8092 Zurich, Switzerland

来源：

JOURNAL OF STATISTICAL PLANNING AND INFERENCE | 1999年 / 79卷 / 01期

关键词：

AR model; context algorithm; discrete nuisance parameter; information criterion; local asymptotic normality; Markov chain; parametric model; semiparametric model; super-efficiency;

D O I：

10.1016/S0378-3758(98)00236-5

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：

We are studying the effect of model selection on parameter estimation in the chosen model. The notion of post-model-selection estimation is used to correctly reflect the fact that the data has been pre-inspected. Under achievable conditions, the post-model-selection estimator is asymptotically equivalent to an efficient estimator where the true model is known. Consequently, it adapts with respect to the model when it is unknown. We give two examples where post-model-selection estimators are efficient and adaptive: one for autoregressive stationary processes, the other for parsimonious stationary Markov chains. (C) 1999 Elsevier Science B.V. All rights reserved.

引用

页码：1 / 9

页数：9

共 17 条

[1]

Basawa IV., 1980, STAT INFERENCE STOCH

[2]

Bickel Peter J, 1993, Efficient and adaptive estimation for semiparametric models, V4

[3]

BUHLMANN P, 1999, IN RPESS ANN STAT

[4]

Drost FC, 1997, ANN STAT, V25, P786

[5]

HALL P, 1980, MARTINGALE LIMIT THE

[6]

HANNAN EJ, 1979, J ROY STAT SOC B MET, V41, P190

[7]

Heyde C.C., 1973, Ann. Probab., V1, P428

[8] ITERATED LOGARITHM RESULT FOR AUTOCORRELATIONS OF A STATIONARY LINEAR PROCESS [J].

HEYDE, CC .

ANNALS OF PROBABILITY, 1974, 2 (02) :328-332

[9] THE EFFECT OF MODEL SELECTION ON CONFIDENCE-REGIONS AND PREDICTION REGIONS [J].

KABAILA, P .

ECONOMETRIC THEORY, 1995, 11 (03) :537-549

[10] Efficient estimation in nonlinear autoregressive time-series models [J].

Koul, HL ;

Schick, A .

BERNOULLI, 1997, 3 (03) :247-277

← 1 2 →