On Model Expansion, Model Contraction, Identifiability and Prior Information: Two Illustrative Scenarios Involving Mismeasured Variables

By Paul Gustafson. 2005

210 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/3939f5a5f2123e7f7c12a7556cabb76a06349d7c

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Statistics and Probability

Citations in the corpus

Penalising Model Component Complexity: A Principled, Practical Approach to Constructing Priors (2014, background): “In these partially identifiable (Gustafson, 2005) or singular (Watanabe, 2009) models, the prior is still present in the limiting posterior.”
Bayesian analysis of measurement error models using integrated nested Laplace approximations (2015, background): “Even if nothing is known and a flat prior is used, Gustafson (2005) has shown that there can be a surprising amount of indirect learning.”
Projecting the future burden of cancer: Bayesian age–period–cohort analysis with integrated nested Laplace approximations (2017, background): “For more details on partially identified models and their asymptotic behavior we refer to Gustafson (2005, 2015).”
Philosophy and the practice of Bayesian statistics (2010, unknown): “Modern Bayesian models in particular are full of unobserved, latent, and unobservable variables, and nonidentifiability is an inevitable concern in assessing such models; see, for example, Gustafson (2005); Vansteelandt et al. (2006); Greenland (2009).”
Relaxation Penalties and Priors for Plausible Modeling of Nonidentified Bias Sources (2009, background): “This use of priors to relax sharp constraints on nonidentified parameters does not entail commitment to Bayesian philosophy, since the resulting penalized estimators can still be evaluated based in part on their frequency properties (Gustafson, 2005; Gustafson and Greenland, 2006, 2010).”
A Framework for Validation of Computer Models (2007, background): “) Gustafson (2005) discussed nonidentifiability in other statistical situations and compared Bayesian and non-Bayesian methods of attempting to deal with the issue, indicating the considerable problems with the latter.”
Bayesian identifiability and misclassification in multinomial data (2004, background): “This is in keeping with one of the basic themes espoused by Gustafson (2004). Having indicated that nonidentifiability can also be problematic in Bayesian analyses, we argue that there are different types of nonidentifiability.”
Bayesian Analysis in Moment Inequality Models (2010, background): “Gustafson (2005) studies Bayesian inference in nonidentified scenarios involving misclassification and measurement errors, which was discussed by a number of prominent researchers.”
What Are the Limits of Posterior Distributions Arising From Nonidentified Models, and Why Should We Care? (2009, methodology): “As a caveat though, this inversion breaks down if the trait prevalence is the same in both populations, and Gustafson (2005a, 2005b) emphasized that correspondingly poor estimation can result if the trait prevalences are similar in both populations.”
Transparent Parametrizations of Models for Potential Outcomes (2012, methodology): “Following Leamer (1978), Gustafson (2005) and Greenland (2005), we argue that partially identified models should be re-parameterized so 570 T. S. Richardson, R. J. Evans and J. M. Robins that the complete parameter vector may be divided into point-identified and entirely non-identified subvectors.”