Bayes without priors

J Clin Epidemiol. 2004 Jan;57(1):4-13. doi: 10.1016/S0895-4356(03)00251-8.

Abstract

Background and objectives: Classical statistical inference has attained a dominant position in the expression and interpretation of empirical results in biomedicine. Although there have been critics of the methods of hypothesis testing, significance testing (P-values), and confidence intervals, these methods are used to the exclusion of all others.

Methods: An alternative metaphor and inferential computation based on credibility is offered here.

Results: It is illustrated in three datasets involving incidence rates, and its advantages over both classical frequentist inference and Bayesian inference, are detailed.

Conclusion: The message is that for those who are unsatisfied with classical methods but cannot make the transition to Bayesianism, there is an alternative path.

Publication types

  • Review

MeSH terms

  • Arizona / epidemiology
  • Bayes Theorem
  • Cluster Analysis
  • Epidemiologic Methods
  • Humans
  • Illinois / epidemiology
  • Incidence
  • Mesothelioma / mortality
  • Precursor Cell Lymphoblastic Leukemia-Lymphoma / epidemiology
  • Statistics as Topic*
  • Survival Rate