×

Diagnostic testing and evaluation of maximum likelihood models. (English) Zbl 0591.62094

Cet article traite d’une methode générale de test de spécification de modèles statistiques. Le cadre est le suivant: \(Y_ 1,..,Y_ n\) sont les observations, i.i.d. Y, v.a., dépendant d’un paramètre \(\theta\). G est la fonction de répartition vraie de Y tandis que \(F(y,\theta)\) est la famille de distributions adoptée pour estimer \(\theta\). \({\hat\theta}_ n\) est l’estimateur du maximum de vraisemblance associé.
Les tests de spécification étudiés ici sont fondés sur des statistiques du type: \({\hat \tau}_ n=n^{-1}\sum^{n}_{1}c(Y_ i,{\hat \theta}_ n)\), les c(y,\(\theta)\) étant des fonctions critères vérifiant (\(\forall \theta):\int c(y,\theta)\quad dF(y,\theta)=0.\)
L’intérêt de la statistique \({\hat\tau}_ n\) est que si le modèle est correctement spécifié, il existe \(\theta_ 0\) tel que \(F'(y,\theta_ 0)\) soit une version de la densité vraie g(y) et par suite: \({\hat \tau}_ n\to^{p.s.}E(c(Y_ i,\theta_ 0))=0.\) Au contraire, en cas d’erreur de spécification, il existe \({\bar\tau}\) (en général \(\neq 0)\) tel que \({\hat \tau}_ n\to^{p.s.}{\bar\tau}.\)
Sous certaines hypothèses est établie la distribution asymptotique du couple (\({\hat \theta}{}_ n,{\hat \tau}_ n)\) ainsi que le comportement local de la limite p.s. \({\bar \tau}\) de \({\hat \tau}{}_ n\) en cas d’erreur de spécification dans une direction v(y) (càd si \(dG(y)=(1+v(y))dF(y,\theta_ 0))\). Un exemple numèrique est fourni, montrant comment, au prix d’un simple test T, on peut pratiquement tester la spécification d’un modèle économétrique.
Reviewer: V.Cohen

MSC:

62P20 Applications of statistics to economics
62F03 Parametric hypothesis testing
62E20 Asymptotic distribution theory in statistics
62H15 Hypothesis testing in multivariate analysis
Full Text: DOI

References:

[1] Aguirre-Torres, V.; Gallant, A. R., The null and non-null asymptotic distribution of the Cox test for multivariate nonlinear regression, Journal of Econometrics, 21, 1-33 (1983) · Zbl 0506.62048
[2] Burguette, J. F.; Gallant, A. R.; Souza, G., On unification of the asymptotic theory of nonlinear econometric models, Econometric Review, 1, 151-190 (1982) · Zbl 0496.62093
[3] Chesher, Andrew, The information matrix test: Simplified calculation via a score interpretation, Economics Letters, 13, 45-48 (1983)
[4] Clarke, Brenton R., Uniqueness and Frechet differentiability of functional solutions to maximum likelihood type equations, Annals of Statistics, 11, 1196-1205 (1983) · Zbl 0541.62023
[5] Cox, D. R., Further results on tests of separate families of hypotheses, Journal of the Royal Statistical Society, B 24, 406-424 (1962) · Zbl 0131.35801
[6] Davidson, R.; Mackinnon, J., Implicity alternatives and the local power of test statistics, Queen’s University discussion paper no. 556 (1984)
[7] Davidson, R.; Mackinnon, J., Model specification tests based on artificial linear regressions, International Economic Review, 25, 485-502 (1984) · Zbl 0554.62099
[8] Engle, Robert F., A general approach to Lagrange multiplier model diagnostics, Journal of Econometrics, 20, 83-104 (1982) · Zbl 0552.62078
[9] Hausman, J., Specification tests in econometrics, Econometrica, 46, 1251-1271 (1978) · Zbl 0397.62043
[10] Heckman, J., The \(χ^2\) goodness of fit for models with parameters estimated from microdata, Econometrica, 52, 1543-1548 (1984) · Zbl 0597.62116
[11] Huber, Peter J., The behavior of maximum likelihood estimates under nonstandard conditions, (LeCam, Lucien M.; Neyman, Jerzy, Proceedings of the fifth Berkeley symposium on mathematical statistics and probability, Vol. 1 (1967), University of California Press: University of California Press Berkeley, CA), 221-234
[12] Huber, Peter J., Robust statistics (1981), Wiley: Wiley New York · Zbl 0536.62025
[13] Kendall, M. G.; Stuart, A., The advanced theory of statistics (1973), Griffin: Griffin London · Zbl 0249.62003
[14] Kiefer, N. and G. Skoog, Local asymtotic specification error analysis, Econometrica 52, 873-886.; Kiefer, N. and G. Skoog, Local asymtotic specification error analysis, Econometrica 52, 873-886. · Zbl 0566.62097
[15] Lancaster, T., The covariance matrix of the information matrix text, Econometrica, 52, 1051-1054 (1984) · Zbl 0564.62097
[16] Moore, D. S.; Spruill, M. C., Unified large-sample of general chi-squared statistics for tests of fit, Annals of Statistics, 3, 599-616 (1975) · Zbl 0322.62047
[17] Newey, W. K., Maximum likelihood specification testing and instrumented score tests (1984), Princeton University mimeo
[18] Phillips, P. C.B., On the consistency of FIML, Econometrica, 50, 1307-1324 (1982) · Zbl 0493.62099
[19] Tauchen, George; Pitts, M., The price variability-volume relationship on speculative markets, Econometrica, 51, 485-505 (1983) · Zbl 0495.90026
[20] White, H., Maximum likelihood estimation of misspecified models, Econometrica, 50, 1-26 (1982) · Zbl 0478.62088
[21] Wouk, A., A course of applied functional analysis (1979), Wiley: Wiley New York · Zbl 0407.46001
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.