Benchmark Priors Revisited: On Adaptive Shrinkage and the Supermodel Effect in Bayesian Model Averaging

By Stefan Zeugner, Martin Feldkircher. Social Science Research Network, 2009

213 citations recorded by the source corpus.

Default prior choices fixing Zellner's g are predominant in the Bayesian Model Averaging literature, but tend to concentrate posterior mass on a tiny set of models. The paper demonstrates this supermodel effect and proposes to address it by a hyper-g prior, whose data-dependent shrinkage adapts posterior model distributions to data quality. Analytically, existing work on the hyper-g-prior is complemented by posterior expressions essential to fully Bayesian analysis and to sound numerical implementation. A simulation experiment illustrates the implications for posterior inference. Furthermore, an application to determinants of economic growth identifies several covariates whose robustness differs considerably from previous results.

Subfields: Economics and Econometrics, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/8075763167a95c3bc8dc8d02ceb4fa9a314e3b68

How researchers see this paper

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

In Economics and Econometrics

In Statistics and Probability

Citations in the corpus

A Default Bayesian Hypothesis Test for ANOVA Designs (2012, background): “(2) 106 Teacher’s Corner It can be seen from Equation (2), and later from Equation (4), that the expected value of g/(g + 1) increases with R2 (Zeugner and Feldkircher 2009).”
Spatial filtering, model uncertainty and the speed of income convergence in Europe (2013, background): “This shape of the posterior probability distribution of β is caused by single speci cations obtaining very large posterior model probabilities, a feature which is related to the use of xed g-priors and has been recently dubbed the supermodel e ect by Feldkircher and Zeugner (2009).”
A variational Bayes approach to variable selection (2017, methodology): “Finally, we compared our method with the Bayesian Model Selection (BMS) method of [21, 22].”
Mixtures of g-Priors in Generalized Linear Models (2015, methodology): “…g, meanwhile leading to asymptotic consistency for model selection and estimation and other attractive theoretical properties (Liang et al. 2008; Feldkircher and Zeugner 2009; Maruyama and George 2011; Bayarri et al. 2012; Celeux et al. 2012; Ley and Steel 2012; Feldkircher 2012; Fouskakis and…”
Market Freedom and the Global Recession (2011, methodology): “As a robustness check, we employ the ’hyper-g prior’ by Liang, Paulo, Molina, Clyde, and Berger (2008) which has been used by Zeugner and Feldkircher (2009) in the context of growth regressions.”
Transmission Lags of Monetary Policy: A Meta-Analysis (2012, methodology): “To address these issues, Bayesian model averaging (BMA) is employed frequently in the literature on the determinants of economic growth (Fernandez et al., 2001; Sala-I-Martin et al., 2004; Durlauf et al., 2008; Feldkircher & Zeugner, 2009; Eicher et al., 2011).”
Does Trust Promote Growth (2013, methodology): “Feldkircher and Zeugner (2009), Ley and Steel (2009) or Eicher et al. (2011) discuss BMA in relation to the determinants of long-term growth.”
Forecast Combination and Bayesian Model Averaging - A Prior Sensitivity Analysis (2012, methodology): “See Feldkircher and Zeugner (2009) for a recent contribution regarding posterior mass ascription and its implications for posterior inference. Furthermore, note that BMA belongs to the class of shrinkage estimators. Under the normal conjugate prior the expected value of the βj is given by E(βj|y,Mj)”
Model Averaging and Its Use in Economics (2017, methodology): “(1997) and Fernández et al. (2001a). On the basis of the application in Masanjala and Papageorgiou (2008), Crespo Cuaresma (2011) finds that MC(3) leads to rather different results from the leaps and bounds method, which does not seem to explore the model space sufficiently well.”