Inferential models: A framework for prior-free posterior probabilistic inference

Abstract

Posterior probabilistic statistical inference without priors is an important but so far elusive goal. Fisher's fiducial inference, Dempster-Shafer theory of belief functions, and Bayesian inference with default priors are attempts to achieve this goal but, to date, none has given a completely satisfactory picture. This paper presents a new framework for probabilistic inference, based on inferential models (IMs), which not only provides data-dependent probabilistic measures of uncertainty about the unknown parameter, but does so with an automatic long-run frequency calibration property. The key to this new approach is the identification of an unobservable auxiliary variable associated with observable data and unknown parameter, and the prediction of this auxiliary variable with a random set before conditioning on data. Here we present a three-step IM construction, and prove a frequency-calibration property of the IM's belief function under mild conditions. A corresponding optimality theory is developed, which helps to resolve the non-uniqueness issue. Several examples are presented to illustrate this new approach.

29 pages with 3 figures. Main text is the same as the published version. Appendix B is an addition, not in the published version, that contains some corrections and extensions of two of the main theorems

References in corpus (9)

• The formal definition of reference priors
• Objective Priors: An Introduction for Frequentists
• Is Bayes Posterior just Quick and Dirty Confidence?
• Calibrated Bayes, for Statistics in General, and Missing Data in Particular
• Marginal inferential models: prior-free probabilistic inference on interest parameters
• Conditional inferential models: combining information for prior-free probabilistic inference
• Dempster--Shafer Theory and Statistical Inference with Weak Beliefs
• Plausibility functions and exact frequentist inference
• Optimal inferential models for a Poisson mean

Cited by in corpus (41)

• Marginal inferential models: prior-free probabilistic inference on interest parameters
• Conditional inferential models: combining information for prior-free probabilistic inference
• False confidence, non-additive beliefs, and valid statistical inference
• Plausibility functions and exact frequentist inference
• A note on p-values interpreted as plausibilities
• On an inferential model construction using generalized associations
• Prior-free probabilistic prediction of future observations
• Frameworks for prior-free posterior probabilistic inference
• Valid inferential models for prediction in supervised learning problems
• Validity, consonant plausibility measures, and conformal prediction
• Direct and approximately valid probabilistic inference on a class of statistical functionals
• Exact prior-free probabilistic inference on the heritability coefficient in a linear mixed model
• Possibility-theoretic statistical inference offers performance and probativeness assurances
• Discussion: Foundations of Statistical Inference, Revisited
• Generalized inferential models for censored data
• Random sets and exact confidence regions
• A possibility-theoretic solution to Basu's Bayesian--frequentist via media
• Inferential models and possibility measures
• Higher order asymptotics of Generalized Fiducial Distribution
• A mathematical characterization of confidence as valid belief
• Asymptotic efficiency of inferential models and a possibilistic Bernstein--von Mises theorem
• Optimal inferential models for a Poisson mean
• Possibilistic inferential models: a review
• Uncertainty Quantification for High Dimensional Sparse Nonparametric Additive Models
• Large-sample theory for inferential models: a possibilistic Bernstein--von Mises theorem
• AutoGFI: Streamlined Generalized Fiducial Inference for Modern Inference Problems in Models with Additive Errors
• Confidence Distribution and Distribution Estimation for Modern Statistical Inference
• Generalized inferential models for meta-analyses based on few studies
• Valid uncertainty quantification about the model in a linear regression setting
• Discussion of "On the Birnbaum Argument for the Strong Likelihood Principle"
• Computationally efficient variational-like approximations of possibilistic inferential models
• Generalized Fiducial Inference for Ultrahigh Dimensional Regression
• A unified nonparametric fiducial approach to interval-censored data
• A fiducial approach to nonparametric deconvolution problem: discrete case
• Partial Conditioning for Inference of Many-Normal-Means with Hölder Constraints
• Approximate Bayesian Computation via Sufficient Dimension Reduction
• A simple recipe for making accurate parametric inference in finite sample
• Exact prior-free probabilistic inference in a class of non-regular models
• A Geometric Perspective on Bayesian and Generalized Fiducial Inference
• Asymptotically optimal inference in sparse sequence models with a simple data-dependent measure
• There is Individualized Treatment. Why Not Individualized Inference?