Prior distributions for variance parameters in hierarchical models (comment on article by Browne and Draper)
Open accessThe DOI matches its Crossref record; 42 of the 42 references it cites were found in a registry or an index; 1 of the 23 sources its claims cite were checked for support.
Abstract
Various noninformative prior distributions have been suggested for scale parameters in hierarchical models. We construct a new folded-noncentral-$t$ family of conditionally conjugate priors for hierarchical standard deviation parameters, and then consider noninformative and weakly informative priors in this family. We use an example to illustrate serious problems with the inverse-gamma family of "noninformative" prior distributions. We suggest instead to use a uniform prior on the hierarchical standard deviation, using the half-$t$ family when the number of groups is small and in other settings where a weakly informative prior is desired. We also illustrate the use of the half-$t$ family for hierarchical modeling of multiple variance parameters such as arise in the analysis of variance.
Overview
CoChat Summary
TL;DR
Gelman argues that inverse-gamma “noninformative” priors for hierarchical variance parameters can badly distort posterior inference, especially when the group-level variance is near zero or the number of groups is small. He recommends a uniform prior on the hierarchical standard deviation when groups are not few, and half-t/half-Cauchy weakly informative priors otherwise, developed within a folded-noncentral-t conditionally conjugate family.
Key findings
- The inverse-gamma(ε, ε) family is not reliably noninformative: as ε → 0 it yields an improper posterior, and in the 8-schools example inverse-gamma(1, 1) and inverse-gamma(0.001, 0.001) priors on σ²_α constrained posterior inferences and increased shrinkage relative to a uniform prior on σ_α.
- A uniform prior on σ_α performed well for the 8-schools data (J = 8), but with only J = 3 schools it produced an extremely long right tail and undershrinkage; a half-Cauchy prior with scale A = 25 removed much of that unrealistic upper tail.
- For multiple variance components, Gelman recommends a hierarchical half-Cauchy model with scale estimated from data; in the 5 × 5 × 2 split-plot latin square Anova with nine variance components, the Cauchy scale parameter A was estimated at 1.8 with a 95% posterior interval of [0.5, 5.1].
- The folded-noncentral-t family, obtained via a redundant multiplicative parameterization, provides conditionally conjugate priors for σ_α and includes the improper uniform density (ν = −1) and proper half-Cauchy (ν = 1) as special cases.
Methods
- Conceptual/comment paper using a simple two-level normal hierarchical model: y_ij ∼ N(μ + α_j, σ²_y), α_j ∼ N(0, σ²_α), focusing on the prior for σ_α.
- Illustrates with the 8-schools example (J = 8, 6000 Bugs simulation draws), a 3-schools subset, and a 5 × 5 × 2 split-plot latin square Anova with nine variance components.
- Uses a parameter-expanded form y_ij ∼ N(μ + ξη_j, σ²_y), η_j ∼ N(0, σ²_η), with σ_α = |ξ|σ_η, to construct and compare uniform, inverse-gamma, half-Cauchy, and hierarchical half-Cauchy priors.
Limitations
- The paper states that noninformative and weakly informative priors are inherently provisional: after fitting, one should inspect the posterior and revise the prior if it does not make sense.
- The paper notes that the uniform prior on σ_α has a positive miscalibration and yields an improper posterior for J = 1 or 2; inverse-gamma inferences are sensitive to ε; and half-Cauchy scale choices are weakly informative rather than fully objective.
- Evident from the text: the recommendations rest on illustrative examples and conceptual argument rather than a broad simulation study or systematic empirical comparison, and extensions to nonlinear, multivariate, and more complex hierarchical models are left for further work.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
0 Valid · 1 Not valid · 14 Not checked · 56 Own results and interpretations
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“Similarly, there are no always-nonnegative classical unbiased estimators of σ α or σ 2 α in the hierarchical model.”
- Also stated in: § 7.1 Prior distributions for variance parameters · plain
“Similarly, there are no alwaysnonnegative classical unbiased estimators of σ α or σ 2 α in the hierarchical model.”
§ Prior distributions for variance parameters in hierarchical models(Comment on Article by Browne and Draper) > 2.4 Calibration
What the sources say
- Could not verifycites Bickel and Blackwell (1967) matched to “A Note on Bayes Estimates” (1967) sourceExperimental — accuracy not yet measured
The introduction (p. 1) and Theorem 1.3 (p. 3) concern when an unbiased Bayes estimate must have zero Bayes risk. Searches for hierarchical models, σₐ, σ²ₐ, and always-nonnegative classical unbiased estimators found no passage addressing the claimed nonexistence result. (LLM, conf=0.77)
- Not checked — no full text of the sourcecites Meng and Zaslavsky (2002) matched to “Single observation unbiased priors” (2002) source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 7.1 Prior distributions for variance parameters · plain
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“If J is low, the uniform prior density tends to lead to high estimates of σ α , as discussed in Section 5.2.”
§ Prior distributions for variance parameters in hierarchical models(Comment on Article by Browne and Draper) > 7.1 Prior distributions for variance parameters
hedgedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In contrast, the hierarchical half-Cauchy model performs a great deal of shrinkage, especially of the high ranges of the intervals.”
- Also stated in: § Bugs model: a half-Cauchy prior distribution on sigma.theta is induced # using a normal prior on xi and an inverse-gamma on tau.eta model { for (j in 1:J){ # J = the number of schools y[j] ~dnorm (theta[j], tau.y[j]) # data model: the likelihood theta[j] <-mu.theta+ xi*eta[j] tau.y[j] <-pow(sigma.y[j],-2) } xi ~dnorm (0, tau.xi) tau.xi <-pow(prior.scale,-2) for (j in 1:J){ eta[j] ~dnorm (0, tau.eta) # hierarchical model for theta } tau.eta ~dgamma (.5, .5)# chi^2 with 1 d.f.sigma.theta<-abs(xi)/sqrt(tau.eta)# cauchy = normal/sqrt(chi^2) mu.theta ~dnorm (0.0, 1.0E-6) # noninformative prior on mu } · plain
“The half-Cauchy model gives much sharper inferences, using the partial pooling that comes with fitting a hierarchical model.”
§ Prior distributions for variance parameters in hierarchical models(Comment on Article by Browne and Draper) > 6.2 Superpopulation and finite-population standard deviations
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § Bugs model: a half-Cauchy prior distribution on sigma.theta is induced # using a normal prior on xi and an inverse-gamma on tau.eta model { for (j in 1:J){ # J = the number of schools y[j] ~dnorm (theta[j], tau.y[j]) # data model: the likelihood theta[j] <-mu.theta+ xi*eta[j] tau.y[j] <-pow(sigma.y[j],-2) } xi ~dnorm (0, tau.xi) tau.xi <-pow(prior.scale,-2) for (j in 1:J){ eta[j] ~dnorm (0, tau.eta) # hierarchical model for theta } tau.eta ~dgamma (.5, .5)# chi^2 with 1 d.f.sigma.theta<-abs(xi)/sqrt(tau.eta)# cauchy = normal/sqrt(chi^2) mu.theta ~dnorm (0.0, 1.0E-6) # noninformative prior on mu } · plain
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“This prior distribution is even more sharply peaked near zero and further distorts posterior inferences, with the problem arising because the marginal likelihood for σ α remains high near zero.”
§ Prior distributions for variance parameters in hierarchical models(Comment on Article by Browne and Draper) > 5.1 Noninformative prior distributions for the 8-schools problem
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The uniform prior distribution seems fine for the 8-school analysis”
- Also stated in: § 5.2 Weakly informative prior distribution for the 3-schools problem · plain
“the default uniform prior gave reasonable results”
- Also stated in: § 7.1 Prior distributions for variance parameters · hedged
“Figure 1 illustrates the generally robust properties of the uniform prior density on σ α .”
§ Prior distributions for variance parameters in hierarchical models(Comment on Article by Browne and Draper) > 5.2 Weakly informative prior distribution for the 3-schools problem
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 5.2 Weakly informative prior distribution for the 3-schools problem · plain
26 more claims
Create a free account to see all 71 claims and every source verdict.
- It may state two separate claims as one. 34
- Probably not a substantial claim. 12
- The quote only partly states this claim. 1
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Because the uniform distribution tends to unduly favor presence of heterogeneity in discrimination and calibration estimates across studies,58,59”
From A framework for meta-analysis of prediction model studies with binary and time-to-event outcomes (2019) · cites this paper as [59]
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Prior distributions for variance parameters in hierarchical models(Comment on Article by Browne and Draper) > 7.1 Prior distributions for variance parameters
“we see conditional conjugacy in the wider class of half-t distributions on σ α , which include the uniform and half-Cauchy densities on σ α (as well as inverse-gamma on σ 2 α ) as special cases.”
§ Prior distributions for variance parameters in hierarchical models(Comment on Article by Browne and Draper) > 3.2 Folded-noncentral-t prior distribution for σ α
“In the limit that the denominator is specified exactly, we have a folded normal distribution; conversely, specifying the numerator exactly yields the square-root-inverse-χ 2 distribution for σ α , as in Section 3.1.”
Citations
Cites
Gelman, A., Carlin, J. B., Stern, H. S., & Rubin, D. B. (2003) · doi:10.1201/9780429258480 · Cited by 27,276
Self-citationKass, R. E., & Raftery, A. E. (1995) · Journal of the American Statistical Association · doi:10.1080/01621459.1995.10476572 · Cited by 12,558
Sampling-based approaches to calculating marginal densities
Gelfand, A. E., & Smith, A. F. M. (1990) · Journal of the American Statistical Association · doi:10.1080/01621459.1990.10476213 · Cited by 6,606 · cited 1×
“Typically, enough data will be available to estimate µ and σy that one can use any reasonable noninformative prior distribution—for example, p(µ, σy) ∝ 1 or p(µ, log σy) ∝ 1.”
Introducing Multilevel Modeling
Kreft, I., & de Leeuw, J. (1998) · doi:10.4135/9781849209366 · Cited by 2,997
POSTERIOR PREDICTIVE ASSESSMENT OF MODEL FITNESS VIA REALIZED DISCREPANCIES
Gelman, A., Meng, X., & Stern, H. S. (1996) · Cited by 2,638
Self-citationBayes and Empirical Bayes Methods for Data Analysis
(2001) · Technometrics · doi:10.1198/tech.2001.s608 · Cited by 2,098
INADMISSIBILITY OF THE USUAL ESTIMATOR FOR THE MEAN OF A MULTIVARIATE NORMAL DISTRIBUTION
Stein, C. (1956) · Contribution to the Theory of Statistics · doi:10.1525/9780520313880-018 · Cited by 1,841
Bayesian Approaches to Clinical Trials and Health‐Care Evaluation
Spiegelhalter, D. J., Abrams, K. R., & Myles, J. P. (2003) · Wiley eBooks · doi:10.1002/0470092602 · Cited by 1,673
Parametric empirical Bayes inference: theory and applications
Morris, C. N. (1983) · Journal of the American Statistical Association · doi:10.1080/01621459.1983.10477920 · Cited by 1,448 · cited 1×
“On the theoretical side, hierarchical models allow a more “objective” approach to inference by estimating the parameters of prior distributions from data rather than requiring them to be specified using subjective information (see James and Stein, 1960, Efron and Morris, 1975, and Morris, 1983).”
The Selection of Prior Distributions by Formal Rules
Kass, R. E., & Wasserman, L. (1996) · Journal of the American Statistical Association · doi:10.1080/01621459.1996.10477003 · Cited by 1,308
Research works printed but not linked: 5
- Box, G. E. P. and Tiao, G. C. (1973). Bayesian Inference in Statistical Analysis. Reading, Mass.: Addison-Wesley.
- James, W. and Stein, C. (1960). "Estimation with quadratic loss." In Neyman, J. (ed.), Proceedings of the Fourth Berkeley Symposium, volume 1, 361–380. Berkeley: University of California Press.
- Johnson, N. L. and Kotz, S. (1972). Distributions in Statistics. New York: Wiley. 4 vols.
- Liu, C. (2001). "Bayesian analysis of multivariate probit models. Discussion of “The art of data augmentation” by D. A. van Dyk and X. L. Meng." Journal of Computational and Graphical Statistics, 10: 75–81.
- Raudenbush, S. W. and Bryk, A. S. (2002). Hierarchical Linear Models. Thousand Oaks, Calif.: Sage., second edition.
Cited by
Items 1,801–1,900 of 4,089
Disruption of Perceptual Learning by a Brief Practice Break
Little, D., Zhang, Y.-X., & Wright, B. A. (2017) · Current Biology · doi:10.1016/j.cub.2017.10.032 · Cited by 12
Residual Normality Assumption and the Estimation of Multiple Membership Random Effects Models
Chen, J., & Leroux, A. J. (2018) · Multivariate Behavioral Research · doi:10.1080/00273171.2018.1533445 · Cited by 12
“increased, there is an active line of research on multilevel sample sizes using Bayesian estimation (cf. Baldwin & Fellingham, 2013; Depaoli, 2013; Gelman, 2006; Hox, van de Schoot, & Matthijsse, 2012; Lambert, Sutton, Burton, Abrams, & Jones, 2005; McNeish, 2016b; McNeish & Stapleton, 2016a; Price, 2012; Soares, Gonç…”
Bayesian copula spectral analysis for stationary time series
Zhang, S. (2018) · Computational Statistics & Data Analysis · doi:10.1016/j.csda.2018.10.001 · Cited by 11
Quick, M., Li, G., & Law, J. (2018) · Geographical Analysis · doi:10.1111/gean.12173 · Cited by 14
Heritability of head size in a hunted large carnivore, the brown bear (Ursus arctos)
Rivrud, I. M., Frank, S. C., Bischof, R., et al. (2019) · Evolutionary Applications · doi:10.1111/eva.12786 · Cited by 11
“These priors follow scaled noncentral F‐ distributions (V = 1 and nu = 1; Gelman, 2006; Hadfield, 2010).”
BackgroundModeling Dependence Structures for Response Times in a Bayesian Framework
Klotzke, K., & Fox, J. (2019) · Psychometrika · doi:10.1007/s11336-019-09671-8 · Cited by 11
Du, H., Edwards, M. C., & Zhang, Z. J. (2019) · Behavior Research Methods · doi:10.3758/s13428-019-01262-w · Cited by 12
“an impact on the resulting posterior distribution and is not a noninformative prior (Gelman, 2006).”
BackgroundBlankers, T., Berdan, E. L., Hennig, R. M., & Mayer, F. (2019) · Evolution · doi:10.1111/evo.13706 · Cited by 11
“We then fitted Bayesian Animal models in MCMCglmm (Hadfield 2010) using an inverse Wishart prior (Gelman 2006) and checked for autocorrelation, effective sample size, and chain convergence following the MCMCglmm course notes (Hadfield 2012).”
MethodBerends, K. D., Scheel, F., Warmink, J. J., de Boer, W. P., Ranasinghe, R., & Hulscher, S. J. M. H. (2019) · Coastal Engineering · doi:10.1016/j.coastaleng.2019.103520 · Cited by 12
White-faced capuchin monkeys use both rank and relationship quality to recruit allies
Kajokaite, K., Whalen, A., Panchanathan, K., & Perry, S. (2019) · Animal Behaviour · doi:10.1016/j.anbehav.2019.06.008 · Cited by 11
Comparison of different software implementations for spatial disease mapping
Vranckx, M., Neyens, T., & Faes, C. (2019) · Spatial and Spatio-temporal Epidemiology · doi:10.1016/j.sste.2019.100302 · Cited by 13
Chakrabarty, A., & Schielzeth, H. (2019) · Heredity · doi:10.1038/s41437-019-0276-1 · Cited by 10
“We used parameter expanded (half Cauchy) priors for our model as they are less informative than regularly used inverse-Wishart priors (Gelman 2006).”
MethodWell-tempered MCMC simulations for population pharmacokinetic models
Bois, F. Y., Hsieh, N., Gao, W., Chiu, W. A., & Reisfeld, B. (2020) · Journal of Pharmacokinetics and Pharmacodynamics · doi:10.1007/s10928-020-09705-0 · Cited by 14
Wang, Z., Fix, M. J., Hug, L., Mishra, A., You, D., Blencowe, H. K., Wakefield, J., & Alkema, L. (2022) · The Annals of Applied Statistics · doi:10.1214/21-aoas1571 · Cited by 13
“We also carry out approximate leave-one-out cross-validation (LOO), which is implemented in the loo package in R (Vehtari, Gelman and Gabry (2017), Vehtari et al.”
BackgroundMethodInfluentialCoppola, P. M., Williams, C., Terhune, T. M., Parke, J. P., & Cecil, J. (2020) · Journal of Wildlife Management · doi:10.1002/jwmg.21975 · Cited by 13
Prediction of Future Failures for Heterogeneous Reliability Field Data
Lewis‐Beck, C., Tian, Q., & Meeker, W. Q. (2021) · Technometrics · doi:10.1080/00401706.2021.1921036 · Cited by 11
Davis, K. P., Augustine, D. J., Monroe, A. P., et al. (2021) · Ornithological Applications · doi:10.1093/ornithapp/duab041 · Cited by 11
Ishitsuka, K., Yamaya, Y., Watanabe, N., et al. (2021) · Geothermics · doi:10.1016/j.geothermics.2021.102316 · Cited by 11
Luan, H., & Fuller, D. (2021) · Environment and Planning B Urban Analytics and City Science · doi:10.1177/23998083211062901 · Cited by 11
Modeling and Simulation of Pretomanid Pharmacodynamics in Pulmonary Tuberculosis Patients
Lyons, M. A. (2019) · Antimicrobial Agents and Chemotherapy · doi:10.1128/aac.00732-19 · Cited by 12
Ogawa, S., Zoda, A., Kagawa, R., et al. (2022) · Theriogenology · doi:10.1016/j.theriogenology.2022.07.005 · Cited by 11
Spatiotemporal connectivity dynamics in spatially structured populations
Drake, J. C., Lambin, X., & Sutherland, C. S. (2022) · Journal of Animal Ecology · doi:10.1111/1365-2656.13783 · Cited by 11
Anticipatory action planning for stepping onto competing potential targets
Watanabe, R., & Higuchi, T. (2022) · Frontiers in Human Neuroscience · doi:10.3389/fnhum.2022.875249 · Cited by 11
“These weak priors were selected independently of the data based on a previous study using hierarchical Bayesian modeling (Gelman, 2006).”
MethodBayes Model Selection with Path Sampling: Factor Models and Other Examples
Dutta, R., & Ghosh, J. K. (2013) · Statistical Science · doi:10.1214/12-sts403 · Cited by 10
Modeling sheep pox disease from the 1994–1998 epidemic in Evros Prefecture, Greece
Malesios, C., Demiris, N., Abas, Z., Dadousis, K., & Koutroumanidis, T. (2014) · Spatial and Spatio-temporal Epidemiology · doi:10.1016/j.sste.2014.07.003 · Cited by 12
“Random year effects (bi) are assumed to be normally distributed with zero mean and their standard deviation following a uniform U(0,100) distribution (Gelman, 2006).”
BackgroundAn analysis of hatching success in the great reed warbler Acrocephalus arundinaceus
Knape, J., Sköld, M., Jonzén, N., Åkesson, M., Bensch, S., Hansson, B., & Hasselquist, D. (2007) · Oikos · doi:10.1111/j.2007.0030-1299.16266.x · Cited by 13
Zhao, L., Shi, J., Shearon, T. H., et al. (2015) · Statistics in Medicine · doi:10.1002/sim.6438 · Cited by 11
Onicescu, G., Lawson, A. B., McDermott, S. W., et al. (2014) · Environmental Science and Pollution Research · doi:10.1007/s11356-014-3072-8 · Cited by 11
“The hyper-parameter standard deviation σβg were assigned a uniform prior on the interval (0,C): σβg∼U 0;Cð Þ; g ¼ 0; 1 (Gelman 2006).”
Thompson, J. A., Bissett, W. T., & Sweeney, A. (2014) · Environmental Health · doi:10.1186/1476-069x-13-47 · Cited by 11
Schulman, D. J., & Bickmore, T. (2010) · Lecture notes in computer science · doi:10.1007/978-3-642-15892-6_14 · Cited by 11
The State, the Economy, and the Regions: Theories of Preference Formation in Times of Crisis
Tatham, M. R., & Bauer, M. W. (2015) · Journal of Public Administration Research and Theory · doi:10.1093/jopart/muv035 · Cited by 16
Qian, X., Nandakumar, R., Glutting, J. J., Ford, D. J., & Fifield, S. (2017) · ETS Research Report Series · doi:10.1002/ets2.12164 · Cited by 12
“Second, the estimation of random variables is more accurate using the fully Bayesian estimation, especially when sample sizes are small (Brown & Draper, 2006; Gelman, 2006). Priors were specified for parameters to be estimated, and they were estimated to posterior distributions. In this study, all parameters were spec…”
MethodHierarchical Bayes approach for subgroup analysis
Hsu, Y., Zalkikar, J. N., & Tiwari, R. C. (2017) · Statistical Methods in Medical Research · doi:10.1177/0962280217721782 · Cited by 13
Unifying and generalizing methods for removing unwanted variation based on negative controls
Gerard, D., & Stephens, M. (2023) · Figshare · doi:10.57912/23894760.v1 · Cited by 11
Conditional Autoregressive ( CAR ) Model
Schmidt, A. M., & Nobre, W. S. (2018) · Wiley eBooks · doi:10.1002/9781118445112.stat08048 · Cited by 18
Hierarchical Shrinkage Priors for Dynamic Regressions with Many PredictorsPreprint
Korobilis, D. (2011) · SSRN Electronic Journal · doi:10.2139/ssrn.1812192 · Cited by 11
On the Importance of Avoiding Shortcuts in Applying Cognitive Models to Hierarchical DataPreprint
Boehm, U., Marsman, M., Matzke, D., & Wagenmakers, E. (2018) · PsyArXiv (OSF Preprints) · doi:10.31234/osf.io/h83kr · Cited by 13
EMC2: An R Package for cognitive models of choicePreprint
Stevenson, N., Donzallaz, M. C., Innes, R. J., Forstmann, B. U., Matzke, D., & Heathcote, A. (2024) · PsyArXiv (OSF Preprints) · doi:10.31234/osf.io/2e4dq · Cited by 18
Variability as a Predictor: A Bayesian Variability Model for Small Samples and Few Repeated MeasuresPreprint
Wiley, J. F., Bei, B., Trinder, J. C., & Manber, R. (2014) · arXiv (Cornell University) · doi:10.48550/arxiv.1411.2961 · Cited by 15
“…second stage outcome use half-Cauchy priors, which has been recommended as a better weakly-informative prior than either the uniform or inverse-gamma families for variance components (Gelman, 2006; Polson & Scott, 2012), specifically with location and scale parameters of zero and ten, respectively.”
MethodA variance shrinkage method improves arm-based Bayesian network meta-analysis
Wang, Z., Lin, L., Hodges, J. S., MacLehose, R. F., & Chu, H. (2020) · Statistical Methods in Medical Research · doi:10.1177/0962280220945731 · Cited by 10
Zhao, Y., Tang, R. S., Du, Y., et al. (2022) · Biometrics · doi:10.1111/biom.13694 · Cited by 10
“Hobbs et al. (2018) usedBayesian predictive probability to evaluate a controlled multiarm platform trial using multiple comparisons. Lin and Bunn (2017) discussed Bayesian multiarm multistage designs with adaptive randomization for platform trials.”
BackgroundMethodInfluentialForecasting Eruptions at Poorly Known Volcanoes Using Analogs and Multivariate Renewal Processes
Wang, T., Bebbington, M., Cronin, S. J., et al. (2022) · Geophysical Research Letters · doi:10.1029/2021gl096715 · Cited by 10
Levey, D. R., Patten, M. A., & Estrada, A. (2022) · Journal of Animal Ecology · doi:10.1111/1365-2656.13871 · Cited by 10
A Correlated Network Scale-Up Model: Finding the Connection Between Subpopulations
Laga, I., Bao, L., & Niu, X. M. (2023) · Journal of the American Statistical Association · doi:10.1080/01621459.2023.2165929 · Cited by 18
Luan, H., Li, G., Duncan, D. T., Sullivan, P. S., & Ransome, Y. (2023) · Annals of Epidemiology · doi:10.1016/j.annepidem.2023.07.004 · Cited by 10
Multi-response robust parameter design based on Bayesian mixed effects model
Chen, X., Wang, J., Ding, C., & Yang, S. (2024) · Applied Mathematical Modelling · doi:10.1016/j.apm.2024.01.008 · Cited by 13
Calibration of burst strength models of corroded pipelines using the hierarchical Bayesian method
Bhardwaj, U., Teixeira, A. P., & Soares, C. G. (2024) · Structural Safety · doi:10.1016/j.strusafe.2024.102444 · Cited by 10
Qian, J., Wu, Y., Zhu, C., et al. (2024) · BMC Public Health · doi:10.1186/s12889-023-17233-y · Cited by 10
“The last term ε it ∼ N ( 0, σ 2 ε ) [28] is the Gaussian random noise variable and captures additional variability not yet explained by other model components.”
BackgroundCoventry, B. S., Lawlor, G. L., Bagnati, C. B., Krogmeier, C., & Bartlett, E. L. (2024) · PNAS Nexus · doi:10.1093/pnasnexus/pgae082 · Cited by 14
A comparison of computational algorithms for the Bayesian analysis of clinical trials
Chen, Z., Berger, J. S., Castellucci, L. A., Farkouh, M. E., Goligher, E. C., Hade, E. M., Hunt, B. J., Kornblith, L. Z., Lawler, P. R., Leifer, E., et al. (2024) · Clinical Trials · doi:10.1177/17407745241247334 · Cited by 14
“45”
Choosing a Proxy Metric from Past Experiments
Tripuraneni, N., Richardson, L. F., D’Amour, A., Soriano, J., & Yadlowsky, S. (2024) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3637528.3671543 · Cited by 32
“[Gelman, 2006] and [Polson and Scott, 2012] both argue for the use of the Half-Cauchy prior for the top-level scale parameter in hierarchical linear models as opposed to the more traditional use of the Inverse-Wishart prior on both empirical and theoretical grounds.”
Stam, M. J., van Wijk, B. C. M., Buijink, A. W. G., Piña‐Fuentes, D., de Bie, R. M. A., Schuurman, R., & Beudel, M. (2025) · npj Parkinson s Disease · doi:10.1038/s41531-025-01053-5 · Cited by 12
“In small datasets such as the one used in this study, the choice of prior can signi fi cantly in fl uence the results(knownaspriorsensitivity) 67 .”
Chu, H., Nie, L., & Kensler, T. W. (2008) · Statistics in Medicine · doi:10.1002/sim.3170 · Cited by 11
Exact prior-free probabilistic inference on the heritability coefficient in a linear mixed model
Cheng, Q., Gao, X., & Martin, R. (2014) · Electronic Journal of Statistics · doi:10.1214/15-ejs984 · Cited by 13
“Bayesian (Gelman, 2006; Gelman et al., 2004; Wolfinger and Kass, 2000) and fiducial methods (Cisewski and Hannig, 2012; E, Hannig and Iyer, 2008; Fisher, 1935) are also available.”
Deng, X. (2014) · Physical review. D. Particles, fields, gravitation, and cosmology/Physical review. D. Particles and fields · doi:10.1103/physrevd.90.024020 · Cited by 11
The Posterior Probability of Passing a Compendial Standard, Part 1: Uniformity of Dosage Units
LeBlond, D. J., & Mockus, L. (2014) · Statistics in Biopharmaceutical Research · doi:10.1080/19466315.2014.928231 · Cited by 10
“If the variance is truly much less than 1.0 however, this prior can highly inflate the posterior distribution and alternative half-Cauchy or half-t priors for variance parameters should be considered (Gelman 2006).”
BackgroundInfluentialPyrogenic carbon distribution in mineral topsoils of the northeastern United States
Jauss, V., Sullivan, P. J., Sanderman, J., Smith, D. B., & Lehmann, J. C. (2017) · Geoderma · doi:10.1016/j.geoderma.2017.02.022 · Cited by 11
“Diffuse priors were used on the parameters following the recommendations of Gelman (2006) and Gelman et al. (2014).”
MethodKorňan, M., Svitok, M., & Krištín, A. (2019) · Ecology and Evolution · doi:10.1002/ece3.5372 · Cited by 10
“We used a half‐Cauchy distribution with a scale parameter of 25, a prior distribution that was recommended as a default tool to handle complete separation in binomial models (Gelman, 2006; Gelman, Jakulin, Pittau, & Su, 2008).”
MethodThe spatial structure of socioeconomic disadvantage: a Bayesian multivariate spatial factor analysis
Quick, M., & Luan, H. (2020) · International Journal of Geographical Information Systems · doi:10.1080/13658816.2020.1759807 · Cited by 11
Sovereign default risk and credit supply: Evidence from the euro area
Palmén, O. (2020) · Journal of International Money and Finance · doi:10.1016/j.jimonfin.2020.102257 · Cited by 18
“3The results are robust to a range of other weakly informative priors between 0.1 and 0.001, which have been used by Gelman (2006) and Jarociński (2010).”
MethodInfluentialBayesian nonparametric quantile process regression and estimation of marginal quantile effects
Xu, S. G., & Reich, B. J. (2021) · Biometrics · doi:10.1111/biom.13576 · Cited by 18
Xu, G., Terada, R., Watanabe, Y., & Nishihara, G. N. (2021) · Journal of Applied Phycology · doi:10.1007/s10811-021-02426-y · Cited by 10
“Informative normal priors were placed on all parameters of the model by using values from a previous study (Kokubu et al. 2015), and a halfCauchy prior distribution was placed on the scale parameter of the models (Gelman 2004, 2006).”
MethodThe Sampling Ratio in Multilevel Structural Equation Models: Considerations to Inform Study Design
Kush, J. M., Konold, T. R., & Bradshaw, C. P. (2021) · Educational and Psychological Measurement · doi:10.1177/00131644211020112 · Cited by 13
“With the appropriate specification of informative priors (e.g., half-Cauchy), Bayesian estimation may be well-suited for MSEM with small sample sizes (see Gelman, 2006; McNeish, 2016).”
MethodQuick, M., & Revington, N. (2021) · Environment and Planning B Urban Analytics and City Science · doi:10.1177/23998083211021419 · Cited by 10
Prior Predictive Checks for the Method of Covariances in Bayesian Mediation Analysis
van Zundert, C., Somer, E., & Miočević, M. (2021) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705511.2021.1977648 · Cited by 10
“001 is considered uninformative for variance parameters in most settings except for multilevel models with very few level-2 units (Gelman, 2006) and structural equation models with small samples (van Erp et al.”
BackgroundMethodInfluentialEstimation of household characteristics with uncertainties from smart meter data
Lin, J., Ma, J., & Zhu, J. G. (2022) · International Journal of Electrical Power & Energy Systems · doi:10.1016/j.ijepes.2022.108440 · Cited by 13
A Bayesian Framework for Multivariate Multifractal Analysis
Leon, L., Wendt, H., Tourneret, J., & Abry, P. (2022) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2022.3187196 · Cited by 14
Detecting Model Misspecification in Bayesian Piecewise Growth Models
Depaoli, S. A., Jia, F., & Heo, I. (2022) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705511.2022.2144865 · Cited by 10
“For example, separation strategy priors can be used for the latent factor covariance matrix (Depaoli, 2021; Liu et al., 2016), and there are a variety of prior settings that can be used for variance parameters (Gelman, 2006).”
MethodEfficient prediction of turbulent flow quantities using a Bayesian hierarchical multifidelity model
Rezaeiravesh, S., Mukha, T., & Schlatter, P. (2023) · Journal of Fluid Mechanics · doi:10.1017/jfm.2023.327 · Cited by 10
Bayesian Model Selection for Generalized Linear Mixed Models
Xu, S., Ferreira, M. A. R., Porter, E. M., et al. (2023) · Biometrics · doi:10.1111/biom.13896 · Cited by 10
“Gelman (2006) 𝜏 has more mass near zero and more mass for large values of 𝜏 than the approximate reference prior for 𝜏 given in Equation (6).”
BackgroundMethodInfluentialPortengen, L., Linet, M. S., Li, G., et al. (2015) · Journal of Exposure Science & Environmental Epidemiology · doi:10.1038/jes.2015.44 · Cited by 10
A Nonparametric Bayesian Learning Model: Application to Text and Image Categorization
Bouguila, N., & Ziou, D. (2009) · Lecture notes in computer science · doi:10.1007/978-3-642-01307-2_42 · Cited by 10
Multivariate spatial hierarchical Bayesian empirical likelihood methods for small area estimation
Porter, A. T., Holan, S. H., & Wikle, C. K. (2015) · Stat · doi:10.1002/sta4.81 · Cited by 10
“Uniform priors are placed on the transformed precision parameters, 1=2 1 and 1=22 , which is equivalent to the practice of placing a flat prior on the standard deviation of a Gaussian process (Gelman, 2006).”
BackgroundRasch models of aphasic performance on syntactic comprehension tests
Gutman, R., DeDe, G., Michaud, J., et al. (2010) · Cognitive Neuropsychology · doi:10.1080/02643294.2010.512285 · Cited by 10
A note on Bayes factor consistency in partial linear models
Choi, T., & Rousseau, J. (2015) · Journal of Statistical Planning and Inference · doi:10.1016/j.jspi.2015.03.009 · Cited by 10
Kim, J.-I., Lawson, A. B., McDermott, S. W., & Aelion, C. M. (2009) · Spatial and Spatio-temporal Epidemiology · doi:10.1016/j.sste.2009.07.003 · Cited by 10
Bayesian variable selection under the proportional hazards mixed-effects model
Lee, K. E., Kim, Y., & Xu, R. (2014) · Computational Statistics & Data Analysis · doi:10.1016/j.csda.2014.02.009 · Cited by 10
“Kinney and Dunson (2007) further considered the parameter-expansion (PX) approach of Gelman (2006) for variance components.”
BackgroundTeaching Bayesian Statistics in a Health Research Methodology Program
Pullenayegum, E., & Thabane, L. (2009) · Journal of Statistics Education · doi:10.1080/10691898.2009.11889537 · Cited by 10
“Background The Gamma(0.001,0.001) prior is the traditional choice for a vague prior for a precision, however Gelman [Gelman, 2006] recently showed that in a variance components model it is in fact informative when the true precision is close to zero (as evidenced by the fact that a Gamma (0.00001,…”
BackgroundInfluentialRanta, J., Lindqvist, R., Hansson, I., et al. (2015) · The Annals of Applied Statistics · doi:10.1214/15-aoas845 · Cited by 10
“One risk factor for human campylobacteriosis is handling and consuming contaminated poultry meat [Kapperud et al. (2003); Wingstrand et al.”
BackgroundInfluentialWildhaber, M. L., Yang, W., & Arab, A. (2014) · River Research and Applications · doi:10.1002/rra.2846 · Cited by 10
“Further, the prior of both variances were specified relatively non-informative inverse-gamma distributionwith shape 1 and scale 1 (Gelman, 2006).”
BackgroundForeman, K. J., Li, G., Best, N., & Ezzati, M. (2016) · Journal of the Royal Statistical Society Series C (Applied Statistics) · doi:10.1111/rssc.12157 · Cited by 11
“D − ρ1A/, where D is diagonal 1/(number of neighbours), A is an adjacency matrix where neighbouring states are 1 and non-neighbours are 0, and τ1 measures the variation across states with weakly informative prior half-Cauchy(0, 5) (Gelman, 2006).”
BackgroundA Bayesian multivariate joint frailty model for disease recurrences and survival
Wen, S., Huang, X., Frankowski, R. F., Cormier, J. N., & Pisters, P. W. T. (2016) · Statistics in Medicine · doi:10.1002/sim.7030 · Cited by 10
“Gelman concluded that very small values in (a, b), corresponding to very high prior variation, are problematic in that inferences can be very sensitive.”
MethodResultInfluentialZachmann, L. J., Shaw, D. W. H., & Dickson, B. G. (2017) · Forest Ecology and Management · doi:10.1016/j.foreco.2017.11.013 · Cited by 11
P-spline smoothing for spatial data collected worldwide
Greco, F. P., Ventrucci, M., & Castelli, E. (2018) · Spatial Statistics · doi:10.1016/j.spasta.2018.08.008 · Cited by 10
“On the topic of prior selection for variance parameters the literature has538 shown rapid growth over the past decade; see, e.g., Gelman (2006); Simpson et al.539 (2017) and references therein.540 The model can be extended in several directions, both in a methodological and ap-541 plied sense.”
BackgroundMethodOn a Robust MaxEnt Process Regression Model with Sample-Selection
Kim, H.-J., Bae, M., & Jin, D. (2018) · Entropy · doi:10.3390/e20040262 · Cited by 10
“See [28], for compatibility with w0 ∼ HC(c, d) to elicit the prior information on w0.”
BackgroundMountaintop removal mining alters stream salamander population dynamics
Price, S. J., Freytag, S. B., Bonner, S. J., Drayer, A. N., Muncy, B. L., Hutton, J. M., & Barton, C. D. (2018) · Diversity and Distributions · doi:10.1111/ddi.12760 · Cited by 12
“The half- t3(2) represents the distribution over the positive real line that is proportional to a t- distribution with three degrees of freedom scaled by a factor of 2 and recommended as a default prior by Gelman (2006).”
BackgroundIntra-Household Heterogeneity and Targeting
Bruno, H. A., Cebollada, J., & Chintagunta, P. K. (2015) · Marketing Science · doi:10.1287/mksc.2018.1088 · Cited by 12
Bayesian Dynamic Linear Models for Estimation of Phenological Events from Remote Sensing Data
Johnson, M. C., Caragea, P. C., Meiring, W., et al. (2018) · Journal of Agricultural Biological and Environmental Statistics · doi:10.1007/s13253-018-00338-y · Cited by 10
“…deviations over the more commonly chosen prior for Bayesian DLMs, the conditionally conjugate inverse-gamma prior on the variances, due to the tendency of the inverse-gamma prior to be informative when a hierarchical variance parameter is small (see, for example, the discussion in Gelman (2006)).”
BackgroundBayesian Modelling of Marked Point Processes with Incomplete Records: Volcanic Eruptions
Wang, T., Schofield, M., Bebbington, M., et al. (2019) · Journal of the Royal Statistical Society Series C (Applied Statistics) · doi:10.1111/rssc.12380 · Cited by 10
“…ns fo r th e re sp ec tiv e V E I ca te go rie s in 20 16 ηi = 315νi1+ exp.−ωi/ , ωi ∼N.μ3, σ 2 3/, .12/ with a standard logistic distribution logistic.0, 1/ as hyperprior for μ3, and a weakly informative half-t prior distribution for the variance parameter, σ3 ∼ t.0, 1, 3/T.0, / (Gelman, 2006).”
Song, Z., Ali, S., & Bouguila, N. (2020) · IET Image Processing · doi:10.1049/iet-ipr.2019.1029 · Cited by 9
Sylvestre, É., Prévost, M., Smeets, P., Medema, G. J., Burnet, J., Cantin, P., Villion, M., Robert, C., & Dorner, S. M. (2020) · Risk Analysis · doi:10.1111/risa.13613 · Cited by 11
“A commonly used prior for the variance parameter of a (log)normal distribution is the half-Cauchy distribution (Gelman, 2006).”
MethodBradley, J. R. (2020) · Bayesian Analysis · doi:10.1214/20-ba1246 · Cited by 15
“Thus, we assume α1 = κ1 = 0, α2 and α3 are distributed according to a gamma distribution, κ2|α2 is distributed according to a shifted (by α2) gamma distribution, κ3 follows a gamma distribution, and v is distributed according to an inverse gamma distribution (e.g., see Gelman, 2006, among others).”
BackgroundLin, C., Hsu, C., Chen, Y., Chang, S., & Gunnell, D. (2021) · Crisis · doi:10.1027/0227-5910/a000793 · Cited by 11
Bayesian global-local shrinkage methods for regularisation in the high dimension linear model
Griffin, J. E., & Brown, P. J. (2021) · Chemometrics and Intelligent Laboratory Systems · doi:10.1016/j.chemolab.2021.104255 · Cited by 10
“In the Bayesian methods, we consider two priors for the observation variance σ 2 : the vague prior π ( σ 2 ) ∝ σ − 2 and the half-Cauchy prior (Gelman, 2006).”
MethodPianori, D., Adja, K. Y. C., Lenzi, J., Pieri, G., Rossi, A., Reno, C., & Fantini, M. P. (2021) · International Journal of Integrated Care · doi:10.5334/ijic.5566 · Cited by 15
Caley, P. A., Reid, J. R. W., Colloff, M. J., & Barry, S. C. (2021) · Environmental and Ecological Statistics · doi:10.1007/s10651-021-00510-7 · Cited by 10
“For priors on variances, following Gelman (2006) we used a uniform prior on the scale of the standard deviation.”
MethodRing, C., Blanchette, A. D., Klaren, W. D., Fitch, S., Haws, L. C., Wheeler, M. W., DeVito, M. J., Walker, N. J., & Wikoff, D. S. (2023) · Regulatory Toxicology and Pharmacology · doi:10.1016/j.yrtph.2023.105464 · Cited by 10
Milanovich, J. R., Hocking, D. J., Peterman, W. E., et al. (2015) · Natural Areas Journal · doi:10.3375/043.035.0412 · Cited by 10
Bayesian Hierarchical Multivariate Formulation with Factor Analysis for Nested Ordinal Data
Savitsky, T. D., & McCaffrey, D. F. (2013) · Psychometrika · doi:10.1007/s11336-013-9339-z · Cited by 10
Singh, M. P., Al‐Yassin, A., & Omer, S. O. (2014) · Crop Science · doi:10.2135/cropsci2014.02.0111 · Cited by 9
“P 2 : the priors for the standard deviation components e , g , r , and b follow a normal distribution N (0, −1 = 10,000) * I (0,) using notation of Gelman (2006).”
BackgroundMethodInfluential
Metadata
- Authors
- Gelman, A.
- Affiliations
- Columbia University (US)
- Corresponding author
- Gelman, Andrew
- ORCID
- Gelman, Andrew
- Venue
- Bayesian Analysis
- Publisher
- Institute of Mathematical Statistics
- Published
- 2006-09-01
- Volume
- 1
- Issue
- 3
- Pages
- 515-534
- ISSN
- 1931-6690 (ISSN-L)
- 1936-0975
- OpenAlex
- W1536497620
- Semantic Scholar
- 63d40805d4635b70f6b5aa07fc2c9821480cfc1e
- Cited by
- 4,050
- References
- 34
- Citations by year
- 2026: 194 · 2025: 218 · 2024: 237 · 2023: 237 · 2022: 282 · 2021: 312 · 2020: 320 · 2019: 281 · 2018: 266 · 2017: 263 · 2016: 235 · 2015: 217 · 2014: 176 · 2013: 142 · 2012: 184
- FWCI
- 78.70
- Type
- journal article
- Language
- en
- Open access
- diamond
- doi.org
- Licence
- cc-by
- Locations
- Bayesian Analysis (publishedVersion, cc-by)
- Project Euclid (Cornell University) (submittedVersion)
- Funders
- National Science Foundation
- Topics
- Physical Sciences › Mathematics › Statistics and Probability › Statistical Methods and Bayesian Inference
- Physical Sciences › Mathematics › Statistics and Probability › Statistical Distribution Estimation and Applications
- Physical Sciences › Computer Science › Artificial Intelligence › Bayesian Methods and Mixture Models
- Keywords
- prior distributions, hierarchical models, scale parameter, non-informative priors, weakly informative priors, uniform prior, analysis of variance
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-02, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-02 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.