×

Imprecision and prior-data conflict in generalized Bayesian inference. (English) Zbl 1211.62051

Summary: A great advantage of imprecise probability models over models based on precise, traditional probabilities is the potential to reflect the amount of knowledge they stand for. Consequently, imprecise probability models promise to offer a vivid tool for handling situations of prior-data conflict in (generalized) Bayesian inference. We consider a general class of recently studied imprecise probability models, including the imprecise Dirichlet model under prior information, and more generally the framework of E. Quaeghebeur and G. de Cooman [Proc. 4th Int. Symp. Imprecise Probab. Appl., 287–296 (2005)] for imprecise inference in canonical exponential families. We demonstrate that such models, in their originally proposed form, prove to be insensitive to the extent of prior-data conflicts. We propose an extension reestablishing the natural relationship between knowledge and imprecision: the higher the discrepancy between the observed sample and what was expected from prior knowledge, the higher the imprecision in the posterior, producing cautious inferences if, and only if, caution is needed. Our approach is illustrated by some examples and simulation results.

MSC:

62F15 Bayesian inference
68T37 Reasoning under uncertainty in the context of artificial intelligence
62F35 Robustness and adaptive procedures (parametric inference)
65C60 Computational problems in statistics (MSC2010)
PDFBibTeX XMLCite
Full Text: DOI

References:

[1] Augustin, T. On the suboptimality of robust Bayesian procedures from the decision theoretic point of view – a cautionary note on updating imprecise previsions. ISIPTA1903: Proc. 3rd Int. Symp. on Imprecise Probabilities and Their Applications. Edited by: Bernard, J. M., Seidenfeld, T. and Zaffalon, M. pp.31–45.
[2] Augustin T., Journal of Statistical Planning and Inference 124 pp 251– (2004) · Zbl 1074.62032 · doi:10.1016/j.jspi.2003.07.003
[3] Bernard J. M., International Journal of Approximate Reasoning 39 pp 123– (2005) · Zbl 1066.62003 · doi:10.1016/j.ijar.2004.10.002
[4] Bernard J. M., International Journal of Approximate Reasoning 50 pp 201– (2009) · doi:10.1016/j.ijar.2008.03.007
[5] Bernardo J. M., Bayesian Theory (1994) · doi:10.1002/9780470316870
[6] Boratyńska A., Statistics & Probability Letters 36 pp 173– (1997) · Zbl 0893.62017 · doi:10.1016/S0167-7152(97)00060-6
[7] Cattaneo M. E.G.V., Ph.D. thesis, in: Statistical Decision Based Directly on the Likelihood Function (2007)
[8] Coolen F. P.A., Statistics & Probability Letters 16 pp 337– (1993) · Zbl 0825.62336 · doi:10.1016/0167-7152(93)90066-R
[9] Coolen F. P.A., The Statistician 43 pp 155– (1994) · doi:10.2307/2348940
[10] Coolen F. P.A., Nonparametric Predictive Inference
[11] Coolen F. P.A., International Journal of Approximate Reasoning (2008)
[12] Cozman, F. G., Nau, R. and Seidenfeld, T. ISIPTA1905: Proceedings of the 4th International Symposium on Imprecise Probabilities and Their Applications. July2005, Pittsburgh, PA, USA. Carnegie Mellon University. www.sipta.orgSIPTA, Manno. See also
[13] de Cooman, G., Vejnarová, J. and Zaffalon, M., eds. ISIPTA1907: Proceedings of the Fifth International Symposium on Imprecise Probabilities and Their Applications. July2007. Prague: Charles University, Czech Republic. SIPTA
[14] Ellsberg D., Q. J. Econ. pp 643– (1961) · Zbl 1280.91045 · doi:10.2307/1884324
[15] Held, H. Quantile-filtered Bayesian learning for the correlation class. ISIPTA1907: Proc. 5th Int. Symp. on Imprecise Probabilities and Their Applications. Edited by: de Cooman, G., Vejnarová, J. and Zaffalon, M. pp.223–232. www.sipta.orgSee also
[16] Held H., International Journal of Approximate Reasoning 49 pp 212– (2008) · Zbl 1184.62037 · doi:10.1016/j.ijar.2008.03.018
[17] Klir G. J., Uncertainty-based information. Elements of generalized information theory (1999) · Zbl 0935.68023
[18] Pericchi L. P., International Statistical Review 58 pp 1– (1991)
[19] Quaeghebeur, E. and de Cooman, G. Imprecise probability models for inference in exponential families. ISIPTA 1905: Proc. 4th Int. Symp. on Imprecise Probabilities and Their Applications. Edited by: Cozman, F. G., Nau, R. and Seidenfeld, T. pp.287–296.
[20] Ríos Insua D., Robust Bayesian Analysis (2000) · Zbl 0958.00015
[21] Troffaes M. C.M., International Journal of Approximate Reasoning 45 pp 17– (2007) · Zbl 1119.91028 · doi:10.1016/j.ijar.2006.06.001
[22] Walley P., Statistical Reasoning with Imprecise Probabilities (1991) · Zbl 0732.62004
[23] Walley P., Journal of the Royal Statistical Society. Series B. Methodological 58 pp 3– (1996)
[24] Walter G., Diploma thesis, in: Robuste Bayes-Regression mit Mengen von Prioris – Ein Beitrag zur Statistik unter komplexer Unsicherheit (2006)
[25] Walter G., Sketch of an alternative approach to linear regression analysis under sets of conjugate priors (2007)
[26] Walter, G., Augustin, T. and Peters, A. Linear regression analysis under sets of conjugate priors. ISIPTA1907: Proc. 5th Int. Symp. on Imprecise Probabilities and Their Applications. Edited by: de Cooman, G., Vejnarová, J. and Zaffalon, M. pp.445–455.
[27] Weichselberger K., International Journal of Approximate Reasoning 24 pp 149– (2000) · Zbl 0995.68123 · doi:10.1016/S0888-613X(00)00032-3
[28] Weichselberger K., Elementare Grundbegriffe einer allgemeineren Wahrscheinlichkeitsrechnung I. Intervallwahrscheinlichkeit als umfassendes Konzept (2001) · Zbl 0979.60001 · doi:10.1007/978-3-642-57583-9
[29] Weichselberger, K. The logical concept of probability and statistical inference. ISIPTA 1905: Proc. 4th Int. Symp. on Imprecise Probabilities and Their Applications. Edited by: Cozman, F. G., Nau, R. and Seidenfeld, T. pp.396–405.
[30] Weichselberger K., Elementare Grundbegriffe einer allgemeineren Wahrscheinlichkeitsrechnung II. Symmetrische Wahrscheinlichkeitstheorie (2009)
[31] Whitcomb K. M., Theory and Decision 58 pp 209– (2005) · Zbl 1087.62040 · doi:10.1007/s11238-005-2458-y
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.