Bayes factors: Prior sensitivity and model generalizability

By Charles Liu, Murray Aitkin. 2008

155 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/cf01f01260ebff5a11db25e9a3d6803b73b54038

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

Default Bayes factors for ANOVA designs (2012, methodology): “There are, however, other Bayesian approaches to inference including Aitkin’s (1991, see Liu & Aitkin, 2008) posterior Bayes factors, Kruschke’s (2010) use of posterior distributions on contrasts, and Gelman and colleagues’ notion of model checking through predictive posterior p-values (e.g.,…”
The philosophy of Bayes’ factors and the quantification of statistical evidence (2016, background): “A number of authors have critiqued the use of Bayes factors for inference on the grounds that they are sensitive to the prior distribution chosen to represent the hypothesis (e.g., Aitkin, 1991; Grünwald, 2000; Liu & Aitkin, 2008; O’Hagan, 1995).”
blavaan: Bayesian structural equation models via parameter expansion (2015, methodology): “The Bayes factor is sensitive to choice of prior distribution (e.g., Liu and Aitkin 2008; Vanpaemel 2010), so researchers using the Bayes factor are advised to carefully consider their priors.”
Bayesian hypothesis testing for psychologists: a tutorial on the Savage-Dickey method. (2010, unknown): “The conceptual challenge is that the Bayesian hypothesis test is acutely sensitive to the specification of the prior distributions for the model parameters (e.g., Bartlett, 1957; Liu & Aitkin, 2008).”
Default Bayes Factors for Model Selection in Regression (2012, background): “For any reasonable data, the probability of the data given a slope as large as 1 million is vanishingly small, and placing weight on these values drives lower the average probability of the data given the model (Lindley, 1957).”
Bayesian Assessment of Null Values Via Parameter Estimation and Model Comparison (2011, unknown): “Thus, the BF for the null model depends strongly on the alternative model to which the null model is compared.4 For an extended discussion with consideration of general model comparison, see Liu and Aitkin (2008).”
Bayesian estimation supersedes the t test. (2013, background): “There are typically a variety of alternative-hypothesis priors, and the relative preference for the null hypothesis can change dramatically depending on the choice of alternative-hypothesis prior (e.g., Dienes, 2008; Kruschke, 2011a; Liu & Aitkin, 2008; Vanpaemel, 2010).”
Bayesian Analysis of Factorial Designs (2017, background): “The fact that Bayes factors are dependent on prior settings has fueled the critique that they are too subjective (e.g., Gelman & Shalizi, 2013; Kruschke, 2013; Liu & Aitkin, 2008).”
Harold Jeffreys’s default Bayes factor hypothesis tests: Explanation, extension, and application in psychology (2016, background): “First, one may object that the test-relevant weighting function2 (i.e., the prior distribution on the parameter of interest) has an overly large influence on the3 Bayes factor (Liu & Aitkin, 2008).”