The Annals of Statistics

Local sensitivity diagnostics for Bayesian inference

Paul Gustafson and Larry Wasserman

Full-text: Open access

Abstract

We investigate diagnostics for quantifying the effect of small changes to the prior distribution over a k-dimensional parameter space. We show that several previously suggested diagnostics, such as the norm of the Fréchet derivative, diverge at rate $n^{k/2}$ if the base prior is an interior point in the class of priors, under the density ratio topology. Diagnostics based on $\phi$-divergences exhibit similar asymptotic behavior. We show that better asymptotic behavior can be obtained by suitably restricting the classes of priors. We also extend the diagnostics to see how various marginals of the prior affect various marginals of the posterior.

Article information

Source
Ann. Statist., Volume 23, Number 6 (1995), 2153-2167.

Dates
First available in Project Euclid: 15 October 2002

Permanent link to this document
https://projecteuclid.org/euclid.aos/1034713652

Digital Object Identifier
doi:10.1214/aos/1034713652

Mathematical Reviews number (MathSciNet)
MR1389870

Zentralblatt MATH identifier
0854.62024

Subjects
Primary: 62F15: Bayesian inference
Secondary: 62F35: Robustness and adaptive procedures

Keywords
Classes of probabilities $\phi$-divergence Fréchet derivatives Kullback-Leibler distance Hellinger distance robustness total variation distance

Citation

Gustafson, Paul; Wasserman, Larry. Local sensitivity diagnostics for Bayesian inference. Ann. Statist. 23 (1995), no. 6, 2153--2167. doi:10.1214/aos/1034713652. https://projecteuclid.org/euclid.aos/1034713652


Export citation

References

  • BASU, S., JAMMALAMADAKA, S. R. and LIU, W. 1996. Local posterior robustness with parametric priors: maximum and average sensitivity. In Maximum Entropy and Bayesian StatisZ. tics G. Heidbreder, ed.. Kluwer, Dordrecht. To appear. Z. Z.
  • BERGER, J. 1984. The robust Bayesian viewpoint with discussion. In Robustness in Bayesian Z. Statistics J. Kadane, ed.. North-Holland, Amsterdam. Z.
  • BERGER, J. 1986. Comment on ``On the consistency of Bay es estimates'' by P. Diaconis and D. Freedman. Ann. Statist. 14 30 37.
  • BERGER, J. 1990. Robust Bayesian analysis: sensitivity to the prior. J. Statist. Plann. Inference 25 303 328. Z.
  • CSISZAR, I. 1977. Information measures: a critical survey. In Transactions of 7th Prague Conference on Information Theory, Statistical Decision Functions and Random Processes 1974. Reidel, Dordrecht. Z.
  • CUEVAS, A. and SANZ, P. 1988. On differentiability properties of Bay es operators. In Bayesian Z. Statistics 3 J. M. Bernardo, M. H. DeGroot, D. V. Lindley and A. F. M. Smith, eds. 569 577. Oxford Univ. Press. Z.
  • DELAMPADY, M. and DEY, D. 1994. Bayesian robustness for multiparameter problems. J. Statist. Plann. Inference 40 375 382. Z.
  • DEROBERTIS, L. 1978. The use of partial prior knowledge in Bayesian inference. Ph.D. dissertation, Yale Univ. Z.
  • DEY, D. and BIRMIWAL, L. R. 1994. Robust Bayesian analysis using entropy and divergence measures. Statist. Probab. Lett. 20 287 294. Z. Z.
  • DIACONIS, P. and FREEDMAN, D. 1986. On the consistency of Bay es estimates with discussion. Ann. Statist. 14 1 67. Z.
  • GELFAND, A. and DEY, D. 1991. On Bayesian robustness of contaminated classes of priors. Statist. Decisions 9 63 80. Z.
  • GOEL, P. 1983. Information measures and Bayesian hierarchical models. J. Amer. Statist. Assoc. 78 408 410. Z.
  • GUSTAFSON, P. 1994. Local sensitivity of posterior expectations. Ph.D. dissertation, Dept. Statistics, Carnegie Mellon Univ. Z.
  • GUSTAFSON, P., SRINIVASAN, C. and WASSERMAN, L. 1995. Local sensitivity analysis. In Bayesian Statistics 5. To appear.Z.
  • KRASKER, W. S. and PRATT, J. W. 1986. Discussion of ``On the consistency of Bay es estimates,'' by P. Diaconis and D. Freedman. Ann. Statist. 14 55 58. Z.
  • LAVINE, M. 1991. An approach to robust Bayesian analysis for multidimensional parameter spaces. J. Amer. Statist. Assoc. 86 400 403. Z.
  • RUGGERI, F. and WASSERMAN, L. 1993. Infinitesimal sensitivity of posterior distributions. Canad. J. Statist. 21 195 203. Z.
  • SIVAGANESAN, S. 1993. Robust Bayesian diagnostics. J. Statist. Plann. Inference 35 171 188. Z.
  • SRINIVASAN, C. and TRUSZCZy NSKA, H. 1990. On the ranges of posterior quantities. Technical Report 294, Dept. Statistics, Univ. Kentucky. Z.
  • WASSERMAN, L. 1992. The conflict between improper priors and robustness. Technical Report 559, Dept. Statistics, Carnegie Mellon Univ.
  • VANCOUVER, BRITISH COLUMBIA PITTSBURGH, PENNSy LVANIA 15213 CANADA V6T 1Z2