A New Calibrated Bayesian Internal Goodness-of-Fit Method: Sampled Posterior p-Values as Simple and General p-Values That Allow Double Use of the Data
2011
A New Method for Goodness-of-Fit Testing in Bayesian Models
publication
Evidence: moderate
Author Information
Author(s): Frédéric Gosselin
Primary Institution: Cemagref, UR EFNO, Nogent-sur-Vernisson, France
Hypothesis
Can a new goodness-of-fit p-value method improve the evaluation of Bayesian statistical models?
Conclusion
The new sampled posterior p-value method provides a more reliable goodness-of-fit assessment than existing methods.
Supporting Evidence
- The new method shows a uniform distribution for p-values under certain conditions.
- Simulations indicate that the new p-value method is stable with less informative priors.
- The method is computationally less intensive compared to other Bayesian goodness-of-fit methods.
Takeaway
This study introduces a new way to check if a statistical model fits the data well, which is easier and more reliable than older methods.
Methodology
The study uses mathematical proofs and simulations to evaluate the performance of the new goodness-of-fit p-value method.
Limitations
The method may not perform well with highly informative priors or in certain specific models.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website