The Coverage Properties of Confidence Regions After Model Selection

作者: Paul Kabaila

DOI: 10.1111/J.1751-5823.2009.00089.X

关键词:

摘要: Summary It is very common in applied frequentist (“classical”) statistics to carry out a preliminary statistical (i.e. data-based) model selection by, for example, using hypothesis tests or minimizing AIC. This usually followed by the inference of interest, same data, based on assumption that selected had been given us a priori. false and it can lead an inaccurate misleading inference. We consider important case interest confidence region. review literature shows resulting regions typically have poor coverage properties. also briefly closely related describes properties prediction intervals after selection. A possible motivation wish utilize uncertain prior information interest. which aim directly construction regions, without requiring intermediate step point this as future direction research. Resume En statistiques appliquees de l'approche frequentiste (“classique”), il est courant proceder une preliminaire du modele statistique (c'est-a-dire basee sur des donnees) en utilisant, par exemple, preliminaires fondes hypotheses ou minimisant Ceci generalement suivi l'inference d'interet, les memes donnees sont utilisees, et qui suppose que le choisi nous avait ete donnea priori. Cette supposition erronee peut entrainer inexacte trompeuse. Nous examinons un cas primordial d'interet constitue region confiance. etudions la documentation indique confiance resultent ont principe proprietes d'application reduites. egalement maniere succincte ecrits etroite relation decrivent intervalles apres statistique. Il sous-tendant represente desir d'utilizer renseignements prealables incertains dans d'interet. l'objectif directement l'elaboration confiance, sans exiger recourir l'etape intermediaire precisons cet objectif axe recherche future.

参考文章(47)
A. K. Md. Ehsanes Saleh, Theory of preliminary test and Stein-type estimation with applications Wiley-Interscience. ,(2006)
Ronald R. Regal, Ernest B. Hook, The effects of model selection on confidence intervals for the size of a closed population. Statistics in Medicine. ,vol. 10, pp. 717- 721 ,(1991) , 10.1002/SIM.4780100506
J. L. Hodges, E. L. Lehmann, The use of Previous Experience in Reaching Statistical Decisions Annals of Mathematical Statistics. ,vol. 23, pp. 47- 58 ,(1952) , 10.1007/978-1-4614-1412-4_6
Paul Chiou, Chien-Pai Han, Conditional interval estimation of the exponential location parameter following rejection of a pre-test Communications in Statistics-theory and Methods. ,vol. 24, pp. 1481- 1492 ,(1995) , 10.1080/03610929508831566
Paul Kabaila, Khreshna Syuhada, The Relative Efficiency of Prediction Intervals Communications in Statistics-theory and Methods. ,vol. 36, pp. 2673- 2686 ,(2007) , 10.1080/03610920701386794
Khageswor Giri, Paul Kabaila, THE COVERAGE PROBABILITY OF CONFIDENCE INTERVALS IN 2rFACTORIAL EXPERIMENTS AFTER PRELIMINARY HYPOTHESIS TESTING Australian & New Zealand Journal of Statistics. ,vol. 50, pp. 69- 79 ,(2008) , 10.1111/J.1467-842X.2007.00500.X
Alexander J. Sutton, Julian P. T. Higgins, Recent developments in meta‐analysis Statistics in Medicine. ,vol. 27, pp. 625- 650 ,(2008) , 10.1002/SIM.2934
Pranab Kumar Sen, Asymptotic Properties of Maximum Likelihood Estimators Based on Conditional Specification Annals of Statistics. ,vol. 7, pp. 1019- 1033 ,(1979) , 10.1214/AOS/1176344785
Borek Puza, Terence O'Neill, Interval estimation via tail functions. Canadian Journal of Statistics-revue Canadienne De Statistique. ,vol. 34, pp. 299- 310 ,(2006) , 10.1002/CJS.5550340207