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 2,701–2,800 of 4,089
Predicting football outcomes and quantifying team strengths with Bayesian modeling
Górecki, T., Grzelak, B., & Dyczkowski, K. (2026) · Expert Systems with Applications · doi:10.1016/j.eswa.2026.131862 · Cited by 3
Yang, H., Chang, F., Muroi, F., Liu, Z., Zhang, W., & Cai, J. (2026) · Psychotherapy and Psychosomatics · doi:10.1159/000552072 · Cited by 3
Reusens, P., & Croux, C. (2017) · Studies in Nonlinear Dynamics and Econometrics · doi:10.1515/snde-2015-0018 · Cited by 8
“In the univariate setting, this type of prior has been used in the Bayesian literature as a non-informative prior for the variance parameter (Gelman, 2006).”
A fast and robust Bayesian update of components failure rates in a nuclear power plant
Ayoub, A., Ariu, V., & Nusbaumer, O. (2019) · Progress in Nuclear Energy · doi:10.1016/j.pnucene.2019.103067 · Cited by 3
Bayesian hierarchical models for service-life prediction of polymers
Pintar, A. L., White, C. C., & Sung, L. (2020) · Elsevier eBooks · doi:10.1016/b978-0-12-818367-0.00011-4 · Cited by 6
Kim, M., Winkler, C. E., Uanhoro, J., Peri, J., & Lochman, J. E. (2021) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705511.2021.1965886 · Cited by 4
“We applied normal priors for coefficients, half-normal priors for level-1 standard deviation and log-normal priors for root variance components (Gelman, 2006) – complete model specification details are in the Appendix A.”
A STUDY OF RESIDENTIAL RADON IN SWEDEN USING MULTI-LEVEL ANALYSIS
Almasri, A. M., Andersson, E. M., & Barregård, L. (2009) · Health Physics · doi:10.1097/01.hp.0000338339.71421.dc · Cited by 3
“Gelman (2006b) recommends starting with a noninformative uniform prior density on standard deviation parameters.”
BackgroundMethodWavelet-L2E Stochastic Volatility Models: an Application to the Water-Energy Nexus
Raath, K. C., & Ensor, K. B. (2022) · Sankhya B · doi:10.1007/s13571-022-00292-3 · Cited by 3
Response surfaces, blocking, and split plots: A predictive distribution case study
Peterson, J. J. (2022) · Quality Engineering · doi:10.1080/08982112.2022.2102427 · Cited by 3
Molinari, M., Cremaschi, A., Iorio, M. D., Chaturvedi, N., Hughes, A. D., & Tillin, T. (2022) · Journal of Applied Statistics · doi:10.1080/02664763.2022.2116746 · Cited by 6
“To improve computational efficiency, we use the representation of the standard half-Cauchy distribution employed by [25,55].”
MethodAvcı, E. (2022) · Turkiye Klinikleri Journal of Biostatistics · doi:10.5336/biostatic.2022-88576 · Cited by 3
Ando, M., IKEDA, T., & Iijima, H. (2023) · Mammal Study · doi:10.3106/ms2021-0049 · Cited by 3
Rehman, H., CHANDRA, N., & Abuzaid, A. (2023) · Hacettepe Journal of Mathematics and Statistics · doi:10.15672/hujms.1066111 · Cited by 3
“Gelman [5] recommended half-t as a default non-informative prior with large and finite value of the variance (scale parameter) of the t distribution.”
BackgroundHierarchical Bayesian propulsion power models — A simplified example with cruise ships
Solonen, A., Maraia, R., Springer, S., et al. (2023) · Ocean Engineering · doi:10.1016/j.oceaneng.2023.115226 · Cited by 3
A Bayesian adaptive design approach for stepped-wedge cluster randomized trials
Wang, J., Cao, J., Ahn, C. W., & Zhang, S. (2024) · Clinical Trials · doi:10.1177/17407745231221438 · Cited by 3
“Gelman 28 showed that the half-Cauchy prior is superior to the inverse-gamma prior for variance parameters, which can severely distort inference.”
BackgroundAlexander-Haw, A., & Breitschopf, B. (2024) · Energy Policy · doi:10.1016/j.enpol.2024.114069 · Cited by 5
New statistical framework for interlaboratory evaluation of anti-doping testing results by WADA
Meija, J., Possolo, A. M., Garrido, B. C., Kisoona, S., & Barroso, O. (2024) · Accreditation and Quality Assurance · doi:10.1007/s00769-024-01595-w · Cited by 4
“…established by WADA in the TD DL [48]: The use of the half-Cauchy distribution to model variance components and the choices for the prior distribution of the laboratory biases are as recommended by [13], widely accepted [30], and adopted and implemented in the NIST Consensus Builder [18].”
Ogawa, R., Engler, J. O., & Cord, A. F. (2024) · Ecological Modelling · doi:10.1016/j.ecolmodel.2024.110778 · Cited by 3
Blouch: Bayesian Linear Ornstein-Uhlenbeck Models for Comparative Hypotheses
Grabowski, M. W. (2024) · Systematic Biology · doi:10.1093/sysbio/syae044 · Cited by 4
“It is important to note that, like any parameter, inspection of the posterior distribution of v may provide new knowledge as to the appropriateness of this prior distribution for further analysis (Gelman 2006).”
Trends in plant cover derived from vegetation plot data using ordinal zero‐augmented beta regression
Strien, A. J. V., Irvine, K. M., & Retel, C. (2024) · Journal of Vegetation Science · doi:10.1111/jvs.13295 · Cited by 3
Palmero, F., Hefley, T. J., Lacasa, J., Almeida, L. F., Haro, R., García, F. O., Salvagiotti, F., & Ciampitti, I. A. (2024) · Plant Methods · doi:10.1186/s13007-024-01261-9 · Cited by 3
“Priors can be selected so that the influence of the prior on the posterior is minimized (which are called weakly informative priors) or based on the prior knowledge about the parameters (informative priors) [28].”
BackgroundMethodKim, J. H., Lawson, A. B., Neelon, B., et al. (2024) · Statistics in Medicine · doi:10.1002/sim.10227 · Cited by 3
Zitzmann, S., Orona, G. A., Lohmann, J. F., König, C., Bardach, L., & Hecht, M. (2024) · Educational and Psychological Measurement · doi:10.1177/00131644241296139 · Cited by 4
“…primarily in research, and scholars have made suggestions about how priors can be specified in advantageous ways (e.g., Chung et al., 2013; Gelman, 2006; Natarajan & Kass, 2000; Smid et al., 2020); see also Zitzmann et al. (2021) who recommended that researchers specify priors in such a…”
Are Bayesian regularization methods a must for multilevel dynamic latent variables models?
Andriamiarana, V., Kilian, P., Brandt, H., & Kelava, A. (2025) · Behavior Research Methods · doi:10.3758/s13428-024-02589-9 · Cited by 4
“We follow recommendations from Gelman (2006) and propose half-Cauchy priors C + ( 0 , a σ ) for standard deviations.”
Energetic cost of human disturbance on the southern sea otter (Enhydra lutris nereis)
Barrett, H., Tinker, M. T., Bentall, G. B., & McDonald, B. I. (2025) · Journal of Wildlife Management · doi:10.1002/jwmg.70012 · Cited by 4
How to add baskets to an ongoing basket trial with information borrowing
Daniells, L., Mozgunov, P., Barnett, H., et al. (2025) · Statistical Methods in Medical Research · doi:10.1177/09622802251316961 · Cited by 3
“The prior on 𝜇 is centred around the average null response rate across the baskets with a large variance, whilst the prior on 𝜎 , g ( ⋅ ) , is more widely debated with inverse-gamma, half-normal or half-Cauchy priors implemented across the literature.”
Cortell‐Nicolau, A., Key, A., & Palomo, A. (2025) · Scientific Reports · doi:10.1038/s41598-025-95054-2 · Cited by 4
“6) (see Gelman 98 , and see Crema & Shoda 108 and Timpson et al. 109 for archaeological applications).”
InfluentialGreen, N., Kurt, M., Moshyk, A., Larkin, J., & Baio, G. (2025) · Statistics in Medicine · doi:10.1002/sim.70132 · Cited by 4
“We considered several alternatives following the recommendations in Gelman (2006) and this was found to be the best option for this study because it balances vagueness and support at zero.”
Liu, H., Heo, I., Ivanov, A., & Depaoli, S. A. (2025) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705511.2025.2503789 · Cited by 5
“The Bayesian inference incorporates prior beliefs for each parameter and updates these beliefs using data collected through Bayes’ theorem (e.g., Gelman, 2006; Kruschke, 2014; Liu et al., 2022): Here, h represents the collection of all model parameters to be estimated, P ð h j data Þ is the…”
Osman, A. A., Esse, A. A., & Muse, A. H. (2025) · BMC Women s Health · doi:10.1186/s12905-025-03900-2 · Cited by 4
“The survival response is created using inla.surv(), and the frailty (random effect) is modeled with an iid structure using the f() function [29, 30].”
Lobo, M. S., & Yao, D. (2025) · Management Science · doi:10.1287/mnsc.2022.04006 · Cited by 3
Bayesian Statistics: A Narrative Review on Application in Inflammatory Bowel Diseases
Camara, H., Vicaut, É., Caron, B., et al. (2025) · Inflammatory Bowel Diseases · doi:10.1093/ibd/izaf148 · Cited by 3
Marchi, M., Grenzi, P., Travascio, A., Uberti, D., Micheli, E. D., Quartaroli, F., Laquatra, G., Pingani, L., Ferrari, S., & Galeazzi, G. M. (2025) · BJPsych Open · doi:10.1192/bjo.2025.10885 · Cited by 7
Mixed Methods for Mixed Models
Dorie, V. (2014) · Columbia Academic Commons (Columbia University) · doi:10.7916/d8v40s5x · Cited by 3
“drawback to their use and echoes the concerns of Gelman (2006). For small values of the parameter, the prior decreases to 0 at a rate faster than that for which the likelihood can compensate.”
BackgroundMethodInfluentialAnalyzing Unreplicated Gauge R&R Studies
Hamada, M. S., & Borror, C. M. (2012) · Quality Engineering · doi:10.1080/08982112.2012.714338 · Cited by 3
“M. S. Hamada and C. M. Borror 544 distributions for random effects models presented in Weaver et al. (2012) that were suggested by Gelman (2006).”
BackgroundMethodAl-Hadhrami, A. Y., & Lawson, A. B. (2010) · Statistical Methods in Medical Research · doi:10.1177/0962280210374256 · Cited by 3
Phase-I design structure of Bayesian variance chart
Saghir, A. (2016) · Cogent Mathematics · doi:10.1080/23311835.2016.1172403 · Cited by 3
“…we have proposed control limits for S2-Chart based on Jeffrey’s prior for unknown mean and variance of the normal distribution, so that we are then relying primarily on the likelihood involved for our inference following the work of Box (1980), Chhikara and Guttman (1982), and Gelman (2006).”
BackgroundMethodInfluentialShared Leadership in Team-Based Learning Classroom Teams and its Relationship to Decision Quality
Peyton, E. J. (2014) · OhioLink ETD Center (Ohio Library and Information Network) · Cited by 3
Bayesian assurance tests for degradation data
Weaver, B., Hamada, M. S., Wilson, A. G., et al. (2017) · Quality and Reliability Engineering International · doi:10.1002/qre.2228 · Cited by 3
Parallel Markov chain Monte Carlo for Bayesian dynamic item response models in educational testing
Wei, Z., Wang, X., & Conlon, E. M. (2017) · Stat · doi:10.1002/sta4.164 · Cited by 3
“Similar methods of centering prior distributions based on data estimates were used in Gelman (2006) and in other application areas including genomics (Tseng et al., 2001; Gottardo et al., 2003; Conlon et al., 2007).”
MethodBollineni-Balabay, O., van den Brakel, J., Palm, F. C., et al. (2017) · Journal of the Royal Statistical Society Series A (Statistics in Society) · doi:10.1111/rssa.12332 · Cited by 3
“(3.9) Setting 𝛼 = 0 (and 𝛾 = 1, without loss of generality) implies that priors on the standard deviation parameters come from the half 𝑡-family, see Gelman (2006).”
BackgroundInfluentialWhither the party vote bonus? Stand-down agreements in Ukraine's mixed-member system
Bloom, S., & Thames, F. C. (2018) · Electoral Studies · doi:10.1016/j.electstud.2018.01.005 · Cited by 3
Han, D. C., & Johnson, K. (2018) · Chemometrics and Intelligent Laboratory Systems · doi:10.1016/j.chemolab.2018.03.015 · Cited by 3
Stevens, J. (2018) · PharmacoEconomics · doi:10.1007/s40273-018-0681-y · Cited by 3
Metric Predicted Variable with Multiple Nominal Predictors
Kruschke, J. K. (2014) · Elsevier eBooks · doi:10.1016/b978-0-12-405888-0.00020-9 · Cited by 3
O'Malley, A. J., James, P., MacKenzie, T. A., Byun, J., Subramanian, S. V., & Block, J. P. (2018) · Statistics in Medicine · doi:10.1002/sim.8039 · Cited by 3
Li, P., Lu, H. C., Gang, Z., Zheng, Q., Yang, L., & Pan, G. (2019) · International World Wide Web Conference (WWW) · doi:10.1145/3308558.3313473 · Cited by 6
“We first augment TSNPF with auxiliary variables to make it conditionally conjugate, which means the complete conditional of each latent variable in p is in the exponential family and is in the same family as its prior [13].”
MethodBayesian influence diagnostics using normalized functional Bregman divergence
Danilevicz, I. M., & Ehlers, R. S. (2020) · Communication in Statistics- Theory and Methods · doi:10.1080/03610926.2020.1764583 · Cited by 3
“…variance structure in which case R ¼ diagonalðr21, :::, r2nÞ: In this case, default prior distributions for the r2i could be Inverse Gamma, while Gamma distributions or even half-t or uniform distributions as suggested by Gelman (2006) might be assigned thus not restricting to conjugate priors.”
BackgroundTang, Y. (2019) · Statistical Methods in Medical Research · doi:10.1177/0962280219865579 · Cited by 5
Kozono, J., Nishihara, G. N., Endo, H., et al. (2019) · Phycological Research · doi:10.1111/pre.12416 · Cited by 3
“Informative normal priors were placed on all parameters of the model by using values from a previous study (Kokubu et al. 2015), and a half-Cauchy prior distribution was placed on the scale parameter of the models (Gelman 2004, 2006).”
MethodHierarchical Bayes Approach to Estimate the Treatment Effect for Randomized Controlled Trials
Liang, X., Kamata, A., & Li, J. (2020) · Educational and Psychological Measurement · doi:10.1177/0013164420909885 · Cited by 5
“Although research on hyper priors has emerged in multilevel modeling (Gelman, 2006a, 2006b), how to formulate hyper priors using past evidence has not been fully addressed and investigated.”
BackgroundMethodInfluentialArevalo, E., Larrañaga, A., & Bardonnet, A. (2020) · Ecology Of Freshwater Fish · doi:10.1111/eff.12546 · Cited by 3
Posterior Averaging Information Criterion
Zhou, S. (2023) · Entropy · doi:10.3390/e25030468 · Cited by 3
Joint analysis of recurrence and termination: A Bayesian latent class approach
Xu, Z., Sinha, D., & Bradley, J. R. (2020) · Statistical Methods in Medical Research · doi:10.1177/0962280220962522 · Cited by 3
Tran, T. D., Lesaffre, E. M. E. H., Verbeke, G. N., & Molenberghs, G. (2020) · Statistics in Medicine · doi:10.1002/sim.8790 · Cited by 3
Bayesian interval-based oncology dose-finding design with repeated quasi-continuous toxicity model
Zhao, D., Zhu, J., & Wang, L. (2021) · Contemporary Clinical Trials · doi:10.1016/j.cct.2021.106265 · Cited by 3
Bayesian Identification of Hamiltonian Dynamics from Symplectic Data
Galioto, N., & Gorodetsky, A. (2020) · IEEE Conference on Decision and Control (CDC) · doi:10.1109/cdc42340.2020.9303852 · Cited by 13
Martins, R. (2021) · Statistical Methods & Applications · doi:10.1007/s10260-021-00566-6 · Cited by 3
Yimer, B. B., & Shkedy, Z. (2021) · RMS Research in Mathematics & Statistics · doi:10.1080/27658449.2021.1896102 · Cited by 3
“Non-informative prior distributions are intended to allow Bayesian inference for parameters about which little is known beyond the data included in the analysis at hand (Gelman, 2006).”
BackgroundMethodInfluentialMultilevel Dynamic Twin Modeling
Schuurman, N. K., Zheng, Y., & Dolan, C. V. (2021) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705511.2021.1937177 · Cited by 3
“For example, a more informative prior distribution that restricts the values that the parameters may take on may improve the estimation for the DZ correlations between the genetic latent variable, or for the random variances in the model MBRWish, Gelman (2006).”
BackgroundSparse linear mixed model selection via streamlined variational Bayes
Degani, E., Maestrini, L., Toczydłowska, D., et al. (2022) · Electronic Journal of Statistics · doi:10.1214/22-ejs2063 · Cited by 3
BNPdensity: Bayesian nonparametric mixture modelling in R
Arbel, J., King, G. K. K., Lijoi, A., Nieto-Barajas, L. E., & Prünster, I. (2021) · Australian & New Zealand Journal of Statistics · doi:10.1111/anzs.12342 · Cited by 4
“Drawing on the flexibility of the algorithm to use non-conjugate prior, we also implemented a range of popular new priors on the scale parameter of the clusters such as the half-Cauchy (Gelman, 2006; Chung et al., 2015), the truncated Gaussian and the uniform distributions.”
BackgroundMethodIs unreliable science guiding bobcat management in Wyoming and other western U.S. states?
Murphy, S. M., Eriksen‐Meier, S., Robertson, L., & Elbroch, L. M. (2022) · Ecological Solutions and Evidence · doi:10.1002/2688-8319.12116 · Cited by 3
Masseloux, J., Le, Q. T., Burr, J., & Gerber, B. D. (2022) · Ecosphere · doi:10.1002/ecs2.3999 · Cited by 4
Kerioui, M., Beaulieu, M., Desmée, S., et al. (2023) · Biometrics · doi:10.1111/biom.13912 · Cited by 3
“All random effects variance priors were half-Cauchy distributions, as recommended when several variance parameters are estimated (Gelman, 2006).”
MethodA note regarding alternative explanations for heterogeneity in meta‐analysis
Senn, S. J., Schmitz, S., Schritz, A., & Araújo, A. (2022) · Statistics in Medicine · doi:10.1002/sim.9403 · Cited by 4
Adaptive Bayesian Sum of Trees Model for Covariate-Dependent Spectral Analysis
Wang, Y., Li, Z., & Bruce, S. A. (2022) · Biometrics · doi:10.1111/biom.13763 · Cited by 4
“A half-t prior is placed on τ b (Gelman, 2006) to complete the Bayesian model specification.”
Verkroost, F. C. J. (2022) · Journal of the Royal Statistical Society Series A (Statistics in Society) · doi:10.1111/rssa.12917 · Cited by 3
“These priors have been recommended as weakly informative priors for usage for SDs in the case of hierarchical models because of shown sharp inference capabilities (Gelman, 2006).”
BackgroundSpanbauer, C., Pan, W., & ADNI, T. A. D. N. I. (2022) · Genetic Epidemiology · doi:10.1002/gepi.22505 · Cited by 2
Informative g-Priors for Mixed Models
Chien, Y.-F., Zhou, H., Hanson, T., & Lystig, T. (2023) · Stats · doi:10.3390/stats6010011 · Cited by 3
Siegel, L. K., & Chu, H. (2023) · Research Synthesis Methods · doi:10.1002/jrsm.1624 · Cited by 3
Phase transition in compressed sensing with horseshoe prior
Nagano, Y., & Hukushima, K. (2023) · Physical review. E · doi:10.1103/physreve.107.034126 · Cited by 3
Sparse Horseshoe Estimation via Expectation-Maximisation
Tew, S. Y., Schmidt, D. F., & Makalic, E. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-26419-1_8 · Cited by 3
A scalable approach for short‐term disease forecasting in high spatial resolution areal data
Orozco‐Acosta, E., Riebler, A., Adín, A., & Ugarte, M. D. (2023) · Biometrical Journal · doi:10.1002/bimj.202300096 · Cited by 5
“Here, we use a uniform improper prior on the positive real line for all standard deviations (square root inverse of precision parameters) in the model (Gelman, 2006; Gómez-Rubio, 2020, Chapter 5.3) and a Uniform [0,1] distribution for the mixing parameter 𝜙 of the BYM2 prior.”
MethodInfluentialZhou, H., Liu, J., Shao, Y., & Tian, Y. (2023) · Educational Studies · doi:10.1080/03055698.2023.2210712 · Cited by 4
“Because of less empirical evidence of homework change points based on the MLPRC model, we chose weakly informative prior distributions for each model by following Gelman’s suggestion (Gelman 2006).”
Huang, L., Hou, Y., Chen, Y., You, X., Proctor, R. W., Francis, G., & Chang, M. (2023) · Applied Ergonomics · doi:10.1016/j.apergo.2023.104048 · Cited by 3
Nolan, T. H., Goldsmith, J., & Ruppert, D. (2023) · Bayesian Analysis · doi:10.1214/23-ba1393 · Cited by 11
“This auxiliary variable-based hierarchical construction facilitates arbitrarily non-informative priors on standard deviation parameters (Gelman, 2006).”
MethodHierarchical Bayesian analysis of racehorse running ability and jockey skills
Nakakita, M., & Nakatsuma, T. (2023) · International Journal of Computer Science in Sport · doi:10.2478/ijcss-2023-0007 · Cited by 4
The Bayesian nested lasso for mixed frequency regression models
Ghosh, S., Khare, K., & Michailidis, G. C. (2023) · The Annals of Applied Statistics · doi:10.1214/22-aoas1718 · Cited by 4
“Another choice is q = 1/p, based on the theoretical results for stochastic regression in Ghosh et al. (2021). For the prior distribution on σ2 ε we follow the suggestion in Gelman et al.”
BackgroundReliability of ordinal outcomes in forensic black-box studies
Arora, H. M., Kaplan‐Damary, N., & Stern, H. S. (2023) · Forensic Science International · doi:10.1016/j.forsciint.2023.111909 · Cited by 3
Proper, J. L., Bunn, V., Hupf, B., & Lin, J. (2023) · Statistics in Biopharmaceutical Research · doi:10.1080/19466315.2023.2292815 · Cited by 3
Deb, S., Roy, R., & Das, S. (2024) · Journal of Forecasting · doi:10.1002/for.3107 · Cited by 3
Two-Stage Limited-Information Estimation for Structural Equation Models of Round-Robin Variables
Jorgensen, T. D., Bhangale, A. M., & Rosseel, Y. (2024) · Stats · doi:10.3390/stats7010015 · Cited by 4
“As advised by Gelman [76] and recommended by the Stan developers (who maintain a web page with recommendations and further reading about specifying priors: https://gith ub.com/stan-dev/stan/wiki/Prior-Choice-Recommendations#prior-for-scale-parameter s-in-hierarchical-models, accessed on 27 January…”
MethodCordero, M., Meinfelder, F., & Eilert, T. (2024) · Journal of Pharmaceutical Sciences · doi:10.1016/j.xphs.2024.02.020 · Cited by 4
Hamiltonian Monte Carlo Sampling for Bayesian Hierarchical Regression in Prognostics
Astfalck, L., & Hodkiewicz, M. R. (2017) · PHM Society Asia-Pacific Conference · doi:10.36001/phmap.2017.v1i1.1821 · Cited by 3
“As (cid:15) → 0 , as to make the distribution non-informative, it yields an improper posterior density, and thus must be set to a reasonable value (Gelman et al., 2006).”
BackgroundIncorporating Subsampling into Bayesian Models for High-Dimensional Spatial Data
Saha, S., & Bradley, J. R. (2024) · Bayesian Analysis · doi:10.1214/24-ba1426 · Cited by 3
“The prior distribution for all the variance parameters (τ 2, σ2, σ2β) are chosen to be a noninformative inverse Gamma distribution, IG(a, b), among different choices (Gelman et al., 1995; Daniels, 1999; Gelman, 2006; Banerjee et al., 2014).”
BackgroundThree Approaches to Improve Inferences Based on Survey Data Collected with Mixed-mode Designs
Yu, W., Elliott, M. R., & Raghunathan, T. E. (2024) · Journal of Survey Statistics and Methodology · doi:10.1093/jssam/smae012 · Cited by 3
de Figueiredo, A., Paterson, P., Lin, L. K., & Mounier‐Jack, S. (2026) · Vaccine · doi:10.1016/j.vaccine.2026.128629 · Cited by 2
Comiskey, H., Alkema, L., & Cahill, N. (2024) · Journal of the Royal Statistical Society Series A (Statistics in Society) · doi:10.1093/jrsssa/qnae051 · Cited by 3
Youm, Y. N., Griffin, J. J., & Bryant, A. P. (2024) · Business Ethics the Environment & Responsibility · doi:10.1111/beer.12720 · Cited by 3
Yorozuya, K., Kubo, Y., Asaoka, Y., Hayashi, H., Fujita, T., & Hanaoka, H. (2024) · Psychogeriatrics · doi:10.1111/psyg.13186 · Cited by 3
“42 MCMC sampling was set to chain 4, iteration 3000, warm-up 2000, and thin 1.”
Ricci, L. E., Cox, M., & Manlove, K. R. (2024) · Philosophical Transactions of the Royal Society B Biological Sciences · doi:10.1098/rstb.2022.0533 · Cited by 3
“We placed a half-t prior distribution of the standard deviation of ram-specific random intercepts and took those intercepts to arise from a normal distribution centred at 0 [50].”
Hierarchical Bayesian Modelling of Macroeconomic Variables in Ghana
Koranteng, E. A., Engmann, G. M., & Dioggban, J. (2024) · Statistics Politics and Policy · doi:10.1515/spp-2024-0013 · Cited by 3
Variance Matrix Priors for Dirichlet Process Mixture Models With Gaussian Kernels
Jing, W., Papathomas, M., & Liverani, S. (2024) · International Statistical Review · doi:10.1111/insr.12595 · Cited by 6
“Gelman (2006) stated that the IW allocates little mass for (1994) pointed out that the IW uses only one degree of freedom ( κ 0 ) to model the variability and dependence between all the parameters in the matrix.”
BackgroundA copula-based Bayesian framework for doping detection
Deliu, N., & Liseo, B. (2024) · Computational Statistics · doi:10.1007/s00180-024-01579-x · Cited by 6
“(6) could be further specified as: where half-t ̄𝜈 is the density of the absolute value of a Student-t random variable centered at zero and with ̄𝜈 degrees of freedom (Gelman 2006), and 𝜋 𝜼 ( 𝓁 ) is a suitable (joint) prior distribution over the family-specific parameters ( 𝓁 ) .”
Stewart, D. R., Sesnie, S. E., Schmidt, P. M., et al. (2025) · Scientific Reports · doi:10.1038/s41598-025-91804-4 · Cited by 3
“This includes the use of latent variables for modeling, data augmentation, and making inferences about shape and scale parameters, which in turn expresses the uncertainty in the posterior probability distribution of the model parameters 61 .”
Aregay, M., Shkedy, Z., & Molenberghs, G. (2014) · Communications in Statistics - Simulation and Computation · doi:10.1080/03610918.2013.781629 · Cited by 3
“The hyper-parameters σ−2θ and σ−2b are assumed to have a flat Gamma prior distribution: σ −2 θ ∼ (0.01, 0.01) and σ−2b ∼ (0.01, 0.01) (Gelman, 2006).”
InfluentialBayesian Spatiotemporal Modeling in Epidemiology: Hepatitis A Incidence Data in Korea
Choi, J. (2014) · Korean Journal of Applied Statistics · doi:10.5351/kjas.2014.27.6.933 · Cited by 3
Bayesian population approaches to the analysis of dose escalation studies
Russu, A., Nicolao, G. D., Poggesi, I., Neve, M., & Goméni, R. (2011) · Computer Methods and Programs in Biomedicine · doi:10.1016/j.cmpb.2011.05.010 · Cited by 3
Sensitivity analysis for variance parameters in Bayesian simplex mixed models for proportional data
Quintero, F. O. L. (2016) · Communications in Statistics - Simulation and Computation · doi:10.1080/03610918.2016.1146763 · Cited by 3
“We are mainly interested in the specification of the prior distribution of the variance of random effect because it has been argued that any Bayesian analysis can be critically dependent on the prior setting (Gelman, 2006, Browne and Draper, 2006).”
Method
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.