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 1,901–2,000 of 4,089
Bayesian variable selection for Poisson regression with underreported responses
Powers, S., Gerlach, R., & Stamey, J. D. (2010) · Computational Statistics & Data Analysis · doi:10.1016/j.csda.2010.04.003 · Cited by 10
MEASURING SUB‐NATIONAL INCOME POVERTY BY USING A SMALL AREA MULTIVARIATE APPROACH
Fabrizi, E., Ferrante, M. R., & Pacei, S. (2008) · Review of Income and Wealth · doi:10.1111/j.1475-4991.2008.00291.x · Cited by 13
Wang, L., & Finn, A. H. R. (2012) · Journal of Targeting Measurement and Analysis for Marketing · doi:10.1057/jt.2012.20 · Cited by 10
Joyce, P. M., Malec, D. J., Little, R. J. A., et al. (2013) · Journal of the American Statistical Association · doi:10.1080/01621459.2013.859077 · Cited by 10
Blending domain estimates from two victimization surveys with possible bias
Lohr, S. L., & Brick, J. M. (2012) · Canadian Journal of Statistics · doi:10.1002/cjs.11153 · Cited by 10
Adaptive Hierarchical Bayes Estimation of Small Area Proportions
Liu, B., & Lahiri, P. (2017) · Calcutta Statistical Association Bulletin · doi:10.1177/0008068317722293 · Cited by 10
“For reference on this prior assumption, we refer to Gelman (2006). We implemented model ~ (0, , ) i v EP σ φ with the two prior assumptions using WinBUGS software. The posterior mean of φ is 0.2. The one-sided 95% credible interval of φ is (0, 0.473), which does not include the normal case ( 0.5 φ = ). The posterior d…”
MethodInfluentialMikkelä, A., Ranta, J., González, M., et al. (2016) · Risk Analysis · doi:10.1111/risa.12572 · Cited by 10
Bayesian joint modelling of the health profile and demand of home care patients
Argiento, R., Guglielmi, A., Lanzarone, E., et al. (2016) · IMA Journal of Management Mathematics · doi:10.1093/imaman/dpw001 · Cited by 10
“Specifically, we assume: Marginal uniform priors for the standard deviation parameters as in (7) or (8) represent a reasonable choice in Bayesian hierarchical models, when a weakly informative prior is used, as in our case (see [21]).”
Oldenkamp, R., Hendriks, H. W. M., van de Meent, D., & Ragas, A. M. J. (2015) · Environmental Science & Technology · doi:10.1021/acs.est.5b02651 · Cited by 12
Small-Area Estimation with Zero-Inflated Data – a Simulation Study
Krieg, S., Boonstra, H. J., & Smeets, M. (2016) · Journal of Official Statistics · doi:10.1515/jos-2016-0051 · Cited by 10
“In SAE it is sometimes useful to include the domain mean xj ¼ 1Nj PNj i¼1xij as auxiliary information (Bafumi and Gelman 2006; Neuhaus and McCulloch 2006).”
MethodHierarchical models for sharing information across populations in phase I dose-escalation studies
Cunanan, K. M., & Koopmeiners, J. S. (2017) · Statistical Methods in Medical Research · doi:10.1177/0962280217703812 · Cited by 10
Brady, S. P., & Goedert, D. (2017) · Copeia · doi:10.1643/cg-16-535 · Cited by 10
“We chose weakly informative proper priors for the random effects (rather than improper, uninformative priors) because the common inverse-gamma and uniform distributions can be sensitive to specifications when the number of groups is small (e.g., two types of environment; Gelman, 2006).”
MethodPleban, J. R., Mackay, D. S., Aston, T. L., Ewers, B. E., & Weinig, C. (2018) · Frontiers in Plant Science · doi:10.3389/fpls.2018.00448 · Cited by 9
“A weakly informed FoldedCauchy distribution (implemented as a truncated t-distribution with one degree of freedom) was used to describe prior distribution for process model variance structure (τ ) (Gelman, 2006).”
MethodKassler, D., Nichols-Barrer, I., & Finucane, M. M. (2019) · Evaluation Review · doi:10.1177/0193841x18818903 · Cited by 10
Simple Penalties on Maximum-Likelihood Estimates of Genetic Parameters to Reduce Sampling Variation
Meyer, K. (2016) · Genetics · doi:10.1534/genetics.115.186114 · Cited by 10
“Discussing priors for variance components in hierarchical models, Gelman (2006) advocated a half-t or half-Cauchy prior with large-scale parameters.”
MethodHeritability and complex segregation analysis of diabetes mellitus in American Eskimo Dogs
Cai, S. V., Famula, T. R., Oberbauer, A. M., et al. (2019) · Journal of Veterinary Internal Medicine · doi:10.1111/jvim.15570 · Cited by 10
“17,19 Finally, the residual term ( e j ) was assumed to be drawn from a standard normal density [i.e., N(0, 1)] as required for this parameterization of a binary trait analysis.”
Variational Bayes for Functional Data Registration, Smoothing, and Prediction
Earls, C., & Hooker, G. (2016) · Bayesian Analysis · doi:10.1214/16-ba1013 · Cited by 16
Characterizing the in-out asymmetry in visual crowding
Chakravarthi, R., Rubruck, J., Kipling, N., & Clarke, A. D. F. (2021) · Journal of Vision · doi:10.1167/jov.21.11.10 · Cited by 10
“…interested in the effect of design condition and their interactions, we assigned a prior distribution of N(0, 1) for each condition. iv) Group-level (random) effects were all assigned a weak half-Cauchy(0, 10) distribution, while half-Cauchy(0, 1) was used for the residual variance (Gelman, 2006).”
MethodClaassen, C., & Traunmüller, R. (2018) · Sociological Methods & Research · doi:10.1177/0049124118769086 · Cited by 12
“Although Gelman (2006) recommends the use of half-Cauchy(0, 2.5) priors for such parameters, independent g distributions can be used to construct a multivariate Dirichlet distribution and are thus preferable when intercepts and slopes vary by group.”
MethodInfluentialGelman, A., van Dyk, D. A., Huang, Z., & Boscardin, W. J. (2007) · Cited by 10
Self-citationMoxham, T., Fox, K. R., Hillsdon, M. M., Green, C., Campbell, J., Foster, C., Mutrie, N., & Searle, J. (2012) · Clinical Governance An International Journal · doi:10.1108/cgij.2012.24817caa.002 · Cited by 10
Deep Evidential RegressionPreprint
Amini, A., Schwarting, W., Soleimany, A. P., & Rus, D. L. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1910.02600 · Cited by 488
“A large focus within Bayesian inference is on placing prior distributions over hierarchical models to estimate uncertainty [12, 13].”
BackgroundUncertainty Shocks in Emerging Economies: A Global to Local Approach for IdentificationPreprint
Miescu, M. (2022) · SSRN Electronic Journal · doi:10.2139/ssrn.4176889 · Cited by 10
“2.2 Prior specification and posterior sampler et al. (2006) shows that results can be sensitive to the choice of the values for s 0 and v 0 and suggest instead the use of a non-informative prior (obtained by setting s 0 = -1 and v 0 = 0) for models where the number of units is greater than 5 which…”
BackgroundInfluentialA Bayesian hierarchical model for discrete choice data in health care
Antonio, A. L., Weiss, R. E., Saigal, C. S., Dahan, E., & Crespi, C. M. (2017) · Statistical Methods in Medical Research · doi:10.1177/0962280217704226 · Cited by 10
The R2D2 Prior for Generalized Linear Mixed Models
Yanchenko, E., Bondell, H. D., & Reich, B. J. (2024) · The American Statistician · doi:10.1080/00031305.2024.2352010 · Cited by 22
“√ Specifically, if W ∼ GBP ( 12 , 12 , 1, σ 2 ) , then W follows a half-Cauchy distribution with scale σ as in Gelman (2006).”
Zhou, Y., Fu, C., Jiang, X., et al. (2024) · Accident Analysis & Prevention · doi:10.1016/j.aap.2024.107752 · Cited by 9
Yang, M., Bao, Q., Shen, Y., et al. (2024) · Analytic Methods in Accident Research · doi:10.1016/j.amar.2024.100361 · Cited by 9
Probing beyond: The impact of model size and prior informativeness on Bayesian SEM fit indices
Edeh, E. J., Liang, X., & Cao, C. (2025) · Behavior Research Methods · doi:10.3758/s13428-025-02609-2 · Cited by 13
“…commonly specified as noninformative prior for variance and covariance parameters, respectively, it is worth noting that these priors have been shown to become unintentionally informative in instances where the variance/covariance parameter is close to zero (Gelman, 2006; Schuurman et al., 2016).”
InfluentialDurability Test and Service Life Prediction Methods for Silicone Structural Glazing Sealant
Yang, B., Liu, J., Li, J., et al. (2025) · Buildings · doi:10.3390/buildings15101664 · Cited by 9
Downscaling and extrapolating dynamic seasonal marine forecasts for coastal ocean users
Vanhatalo, J. P., Hobday, A. J., Little, L. R., & Spillman, C. M. (2016) · Ocean Modelling · doi:10.1016/j.ocemod.2016.01.004 · Cited by 8
Propagation of uncertainties modelled by parametric p-boxes using sparse polynomial chaos expansions
Schöbi, R., & Sudret, B. (2015) · Repository for Publications and Research Data (ETH Zurich) · doi:10.3929/ethz-a-010494112 · Cited by 9
“…the latter lack of knowledge (epistemic uncertainty) as well as the natural variability of the process (aleatory uncertainty), including probability-boxes (Ferson and Ginzburg, 1996), Bayesian hierarchical models (Gelman, 2006) and Dempter-Shafer’s evidence theory (Dempster, 1967; Shafer, 1976).”
InfluentialA Bayesian hierarchical approach for spatial analysis of climate model bias in multi-model ensembles
Arisido, M. W., Gaetan, C., Zanchettin, D., & Rubino, A. (2017) · Stochastic Environmental Research and Risk Assessment · doi:10.1007/s00477-017-1383-2 · Cited by 9
“…we only 507 focus on θ as hyperparameters of variance components are 508 more sensitive than the hyperparameters of other forms such509 as Gaussian prior for regression coefficients (Gelman 2006).510 Figure 8 illustrates the posterior distributions of τ j for the511 three choices θ ∈ {20,35,40}.”
BackgroundMethodResultA hierarchical model for the ages of Galactic halo white dwarfs
Si, S., van Dyk, D. A., von Hippel, T., Robinson, E., Webster, A., & Stenning, D. C. (2017) · Monthly Notices of the Royal Astronomical Society · doi:10.1093/mnras/stx765 · Cited by 11
“In statistics, hierarchical models are viewed as the most efficient and principled technique to estimate the parameters of each object in a population (e.g., Gelman 2006).”
MethodWoods, P. J., & Jónasson, J. P. (2017) · Fisheries Research · doi:10.1016/j.fishres.2017.05.011 · Cited by 9
Effects of prescribed fire on San Francisco gartersnake survival and movement
Halstead, B. J., Thompson, M. E., Amarello, M., Smith, J. J., Wylie, G. D., Routman, E. J., & Casazza, M. L. (2018) · Journal of Wildlife Management · doi:10.1002/jwmg.21585 · Cited by 9
Random main effects of treatment: A case study with a network meta‐analysis
Senn, S. J., Schmitz, S., Schritz, A., & Salah, S. (2019) · Biometrical Journal · doi:10.1002/bimj.201700265 · Cited by 9
Content Popularity Estimation in Edge-Caching Networks from Bayesian Inference Perspective
Mehrizi, S., Tsakmalis, A., Chatzinotas, S., & Ottersten, B. (2019) · IEEE Consumer Communications and Networking Conference · doi:10.1109/ccnc.2019.8651737 · Cited by 8
“As it is discussed in [16], Half-Cauchy distribution performs well in a training set with small number of samples as a non-informative prior.”
BackgroundComparing Goal-Based and Result-Based Approaches in Modelling Football Outcomes
Egidi, L., & Torelli, N. (2020) · Social Indicators Research · doi:10.1007/s11205-020-02293-z · Cited by 9
Rawson, T., Paton, R. S., Colles, F. M., et al. (2020) · Frontiers in Microbiology · doi:10.3389/fmicb.2020.576646 · Cited by 9
“Following the advice of Gelman (2006) for non-informative improper priors, we use a uniform distribution between 0 and 50 for the prior distribution of each of these standard deviation parameters.”
MethodLloyd-Smith, P. R., Adamowicz, W., Entem, A., Fenichel, E. P., & Rad, M. R. (2021) · Journal of Economic Behavior & Organization · doi:10.1016/j.jebo.2020.12.009 · Cited by 9
Kinlock, N. L., & Munch, S. B. (2021) · Oikos · doi:10.1111/oik.08453 · Cited by 9
Bayesian quantile regression for longitudinal count data
Jantre, S. R. (2022) · Journal of Statistical Computation and Simulation · doi:10.1080/00949655.2022.2096025 · Cited by 9
Witczak, L. R., Blozis, S. A., & Bales, K. L. (2022) · Animal Behaviour · doi:10.1016/j.anbehav.2022.07.001 · Cited by 11
“The random intercept was assumed to be normally distributed, and given the small sample at the pair level, the standard deviation of the random intercept was assumed to follow a half-t distribution with 3 degrees of freedom and to have a lower bound of 0 and a variance of 10 (see Gelman, 2006).”
MethodCharacterization of the highly fractured zone at the Grimsel Test Site based on hydraulic tomography
Ringel, L. M., Jalali, M., & Bayer, P. (2022) · Hydrology and earth system sciences · doi:10.5194/hess-26-6443-2022 · Cited by 9
Sample size calculation for clinical trials analyzed with the meta‐analytic‐predictive approach
Qi, H., Rizopoulos, D., & van Rosmalen, J. (2023) · Research Synthesis Methods · doi:10.1002/jrsm.1618 · Cited by 10
Wang, S., Ren, Z., & Liu, X. (2023) · Frontiers in Public Health · doi:10.3389/fpubh.2023.996694 · Cited by 16
Sánchez-Céspedes, L. M., Leasure, D. R., Tejedor‐Garavito, N., et al. (2023) · Population Studies · doi:10.1080/00324728.2023.2190151 · Cited by 9
“We chose uniform priors for standard deviations rather than the half-Cauchy priors suggested by Gelman (2006), to avoid a long tail that included unrealistic parameter space on the log scale.”
Petrovska, R., Bugmann, H. K. M., Hobi, M. L., & Brang, P. (2023) · Forest Ecology and Management · doi:10.1016/j.foreco.2023.120939 · Cited by 13
Odor identification errors reveal cognitive aspects of age-associated smell loss
Raj, R., Hörberg, T., Lindroos, R., Larsson, M. U., Herman, P. A., Laukka, E. J., & Olofsson, J. (2023) · Cognition · doi:10.1016/j.cognition.2023.105445 · Cited by 11
“We used standard weakly regularizing priors (Gelman, 2006, p. 57; Gelman & Hill, 2006; Stan Development Team, 2017).”
Limited host availability disrupts the genetic correlation between virulence and transmission
Godinho, D. P., Rodrigues, L. R., Lefèvre, S., et al. (2023) · Evolution Letters · doi:10.1093/evlett/qrac008 · Cited by 9
“002), which gave similar results (not shown) (Gelman, 2006).”
BackgroundMethodResultInfluentialGo Multivariate: Recommendations on Bayesian Multilevel Hidden Markov Models with Categorical Data
Moraga, S. M., & Aarts, E. (2023) · Multivariate Behavioral Research · doi:10.1080/00273171.2023.2205392 · Cited by 18
“The overestimation is likely a consequence of using a scaled Inverse-Wishart distribution to sample the variance-covariance matrices of random effects, which has a low density around zero (Gelman, 2006; Lemoine, 2019).”
MethodRevealing Senior Mobility Patterns and Activities in Urban Transit Systems
Chen, E., Liu, Y., & Yang, M. (2023) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2023.3275389 · Cited by 9
Wolfhagen, J. (2023) · Journal of Archaeological Method and Theory · doi:10.1007/s10816-023-09611-y · Cited by 8
“Several approaches exist for deciding how to describe this prior belief, ranging from “objective” priors that provide equal weight to all possible values of a parameter to distinct distributions defined by a synthesis of previous or related research (Gelman, 2006b).”
Chau, J., Altan, S., Burggraeve, A., Coppenolle, H., Kifle, Y. W., Prokopcová, H., Daele, T. V., & Sterckx, H. (2023) · AAPS PharmSciTech · doi:10.1208/s12249-023-02695-5 · Cited by 13
“F o r t h e noise standard deviations e and , we assign a Half-Student-t ( 𝜈 = 5, 𝜇 = 0, 𝜎 = 1 ) prior distribution, recommended as a weakly informative prior for scale parameters in [37].”
Awareness of embodiment enhances enjoyment and engages sensorimotor cortices
Moffat, R., & Cross, E. S. (2024) · Human Brain Mapping · doi:10.1002/hbm.26786 · Cited by 13
“…priors to impose a constrained distribution on our expected results, as a means of acknowledging our existing knowledge regarding the expected outcomes, allowing for plausible large effects, and allowing the data to dominate the structure of the posterior distribution (Gelman, 2006; Lemoine, 2019).”
BackgroundUsing degradation models to assess pipeline life
Tian, Q., Liu, S., & Meeker, W. Q. (2019) · Applied Stochastic Models in Business and Industry · doi:10.1002/asmb.2489 · Cited by 9
“The parameters for Model 2 in (2) are: θ2 = log(β11) , log(β12) , log(β13) , log(β14) , log(β11) , log(β12) , log(β13) , log(β14) , ρ12, ρ13, ρ14, ρ23, ρ24, ρ34, σ ) ′.”
BackgroundMethodResultInfluentialCorberán‐Vallet, A., & Gómez, F. J. S. (2014) · Biometrical Journal · doi:10.1002/bimj.201300194 · Cited by 9
“All the standard deviations in the previous equations are assigned the Uniform distribution in the interval (0,5) (Gelman, 2006).”
BackgroundWikström, A., Knape, J., Casini, M., et al. (2015) · Population Ecology · doi:10.1007/s10144-015-0520-3 · Cited by 9
“…normal priors with mean 0 and variance 10,000, the standard deviations of the process and observation errors, rp and rm, were given half-t5 priors (Gelman 2006) and the logarithm of the initial population state loge(J1) was given a normal prior with mean equal to the first observed index value…”
MethodBSMac: A MATLAB toolbox implementing a Bayesian spatial model for brain activation and connectivity
Zhang, L., Agravat, S. B., Derado, G., Chen, S., McIntosh, B. J., & Bowman, F. D. (2011) · Journal of Neuroscience Methods · doi:10.1016/j.jneumeth.2011.10.025 · Cited by 9
“However, these assumptions can lead to poor estimation in some settings (Natarajan and Kass, 2000; Kass and Natarajan, 2006; Gelman, 2006).”
BackgroundAregay, M., Shkedy, Z., & Molenberghs, G. (2012) · Computational Statistics & Data Analysis · doi:10.1016/j.csda.2012.06.020 · Cited by 9
“To complete the specification of the hierarchical model, assume the following hyper-prior distributions: σ−2b ∼ G(0.01, 0.01) and σ −2 βlk ∼ G(0.01, 0.01) Non-informative independent normal priors µlk ∼ N(0; 100, 000) were specified for the prior means of the components of β (Gelman 2006).”
BackgroundInfluentialBayesian latent variable models for spatially correlated tooth-level binary data in caries research
Zhang, Y., Todem, D., Kim, K., & Lesaffre, E. M. E. H. (2011) · Statistical Modelling · doi:10.1177/1471082x1001100103 · Cited by 9
“…spatial variability parameter σ j , j = 1, . . . , 4, three different prior specifications were considered: σ −2 j ∼ Statistical Modelling 2011; 11(1): 25–47 IG(0.001, 0.001)(Spiegelhalter et al., 2003), σ 2j ∼ IG(0.5, 0.005)(GeoBUGS: Thomas et al., 2004) or σ j ∼ U(0, 1000) (Gelman 2003, 2006).”
BackgroundInfluentialBayesian multiple comparisons of simply ordered means using priors with a point mass
Nashimoto, K., & Wright, F. T. (2008) · Computational Statistics & Data Analysis · doi:10.1016/j.csda.2008.05.015 · Cited by 9
Testing volatility persistence on Markov switching stochastic volatility models
Pan, Q., & Li, Y. (2013) · Economic Modelling · doi:10.1016/j.econmod.2013.06.029 · Cited by 9
Boucher, M. (2012) · Pharmaceutical Statistics · doi:10.1002/pst.1515 · Cited by 9
Intralocus sexual conflict, adaptive sex allocation, and the heritability of fitness
Calsbeek, R. G., Duryea, M. C., Goedert, D., et al. (2015) · Journal of Evolutionary Biology · doi:10.1111/jeb.12713 · Cited by 9
“A parameter-extended prior takes the form of a noncentral F-distribution, as described by Gelman (2006), and is recommended to account for near-zero variances (Hadfield, 2014).”
MethodPita, R., Lambin, X., Mira, A. P. P., et al. (2016) · Oecologia · doi:10.1007/s00442-016-3653-y · Cited by 9
Chamberlain, D. B., & Chamberlain, J. M. (2016) · Annals of Emergency Medicine · doi:10.1016/j.annemergmed.2016.08.449 · Cited by 10
Sheng, Y. (2017) · Frontiers in Psychology · doi:10.3389/fpsyg.2017.00123 · Cited by 9
“Gelman (2006), however, in an attempt to illustrate the performance of these prior distributions for σ near zero, which is where classical and Bayesian inferences differ the most (see Brown and Draper, 2006), pointed out problems with the latter two, especially with the inverse-gamma family of…”
BackgroundMethodInfluentialKivelä, S. M., Friberg, M., Wiklund, C., & Gotthard, K. (2017) · Biological Journal of the Linnean Society · doi:10.1093/biolinnean/blx047 · Cited by 9
Bayesian Approach to Utilize Historical Control Data in Clinical Trials
Takeda, K., Oba, M. S., Kakizume, T., Sakamaki, K., Taguri, M., & Morita, S. (2015) · Japanese Journal of Biometrics · doi:10.5691/jjb.36.25 · Cited by 10
A multilevel Bayesian approach for modeling the time-to-serve in professional tennis
Kovalchik, S. A., & Albert, J. (2017) · Journal of Quantitative Analysis in Sports · doi:10.1515/jqas-2016-0091 · Cited by 9
“The interested reader can read about different choices of noninformative priors for the secondstage variance parameters in multilevel models (Gelman 2006).”
BackgroundNakamura, Y. (2017) · Explorations in Economic History · doi:10.1016/j.eeh.2017.06.001 · Cited by 9
Drivers of population growth variations for two Mediterranean sympatric Deer
López‐Montoya, A. J., Moro, J., & Azorit, C. (2017) · Ecological Research · doi:10.1007/s11284-017-1486-8 · Cited by 9
Zero‐inflated multiscale models for aggregated small area health data
Aregay, M., Lawson, A. B., Faes, C., et al. (2017) · Environmetrics · doi:10.1002/env.2477 · Cited by 9
“For this and the subsequent models, we use noninformative normal prior distributions for the “fixed” effect parameters, whereas we employ uniform prior distributions for the standard deviations of the random effects (Gelman, 2006).”
MethodA Science and Risk-Based Pragmatic Methodology for Blend and Content Uniformity Assessment
Sayeed-Desta, N., Pazhayattil, A. B., Collins, J., et al. (2017) · AAPS PharmSciTech · doi:10.1208/s12249-017-0925-8 · Cited by 9
“If a more precise estimate is required, the recommendation for performing VCA with 12 location triplicates can be followed (14).”
BackgroundPina-Sánchez, J., Koskinen, J. H., & Plewis, I. (2019) · Journal of Official Statistics · doi:10.2478/jos-2019-0010 · Cited by 9
Bayesian forecasting of electoral outcomes with new parties’ competition
Montalvo, J. G., Papaspiliopoulos, O., & Stumpf-Fétizon, T. (2019) · European Journal of Political Economy · doi:10.1016/j.ejpoleco.2019.01.006 · Cited by 9
“Bending” and beyond: Better estimates of quantitative genetic parameters?
Meyer, K. (2019) · Journal of Animal Breeding and Genetics · doi:10.1111/jbg.12386 · Cited by 11
Ethical and Statistical Considerations in Models of Moral Judgments
Sivill, T. (2019) · Frontiers in Robotics and AI · doi:10.3389/frobt.2019.00039 · Cited by 12
“Prior specification becomes particularly challenging when using hierarchical models for Bayesian inference as these models require hyperparameters, each requiring a prior distribution (Gelman, 2006).”
BackgroundInfluentialPereira, J. B., Nobre, W. S., Silva, I. F. L., & Schmidt, A. M. (2020) · Journal of the Royal Statistical Society Series A (Statistics in Society) · doi:10.1111/rssa.12551 · Cited by 10
“For the variance parameters σ2 and τ2, we assign independent half-Cauchy prior distributions to the standard deviations (Gelman, 2006).”
MethodProvider Contributions to Disparities in Mental Health Care
Merced, K., Imel, Z. E., Baldwin, S. A., Fischer, H. J., Yoon, T., Stewart, C. C., Simon, G., Ahmedani, B. K., Beck, A. L., Daida, Y. G., et al. (2020) · Psychiatric Services · doi:10.1176/appi.ps.201800500 · Cited by 13
Tran, T. D., Lesaffre, E. M. E. H., Verbeke, G. N., & Duyck, J. A. J. (2020) · Biometrics · doi:10.1111/biom.13292 · Cited by 12
Nonparametric hierarchical mixture models based on asymmetric Gaussian distribution
Song, Z., Ali, S., Bouguila, N., & Fan, W. (2020) · Digital Signal Processing · doi:10.1016/j.dsp.2020.102829 · Cited by 9
Uncertainty evaluations from small datasets
Stoudt, S., Pintar, A. L., & Possolo, A. M. (2020) · Metrologia · doi:10.1088/1681-7575/abd372 · Cited by 11
Buffington, T., Scott, J. G., & Ezekoye, O. A. (2021) · Computers Environment and Urban Systems · doi:10.1016/j.compenvurbsys.2021.101633 · Cited by 12
Wittman, J. T., Silk, P. J., Parker, K., & Aukema, B. H. (2021) · Agricultural and Forest Entomology · doi:10.1111/afe.12457 · Cited by 9
Fisher’s z Distribution-Based Mixture Autoregressive Model
Solikhah, A., Kuswanto, H., Iriawan, N., et al. (2021) · Econometrics · doi:10.3390/econometrics9030027 · Cited by 9
“(Gelman 2006) Gelman, Andrew.”
MethodPicón, A., Álvarez-Gila, A., Arteche, J. A., et al. (2021) · IEEE Access · doi:10.1109/access.2021.3096599 · Cited by 9
“And the standard deviation of the distribution is also a random variable with a prior following a Half-Cauchy distribution [38] with a fixed beta parameter value:”
MethodKeogan, K., Daunt, F., Wanless, S., Phillips, R. A., Álvarez, D., Anker‐Nilssen, T., Barrett, R. T., Bech, C., Becker, P. H., Berglund, P., et al. (2022) · Journal of Animal Ecology · doi:10.1111/1365-2656.13758 · Cited by 12
“To improve mixing, for the remaining variance random terms, we ad - opted parameter-expanded priors (Gelman et al., 2008 ), which give a scaled F distribution with numerator and denominator degrees of freedom = 1 and scale parameter = 1,000 (Gelman, 2006 ).”
MethodAbkarian, H., Hegde, S. J., & Mahmassani, H. S. (2022) · Transportation Research Record Journal of the Transportation Research Board · doi:10.1177/03611981221098665 · Cited by 10
“The prior corresponds to a flat prior on m and t, and is recommended by Gelman (28).”
BackgroundStrader, M. E., Wolak, M. E., Simon, O. M., & Hofmann, G. E. (2022) · Proceedings of the Royal Society B Biological Sciences · doi:10.1098/rspb.2022.1249 · Cited by 14
Bayesian Traffic State Estimation Using Extended Floating Car Data
Kyriacou, V., Englezou, Y., Panayiotou, C. G., & Timotheou, S. (2022) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2022.3225057 · Cited by 11
“10The particular values were chosen to ensure an uninformative prior, as well as to eliminate any issues in the posterior inference [68].”
Territory and population attributes affect Florida scrub‐jay fecundity in fire‐adapted ecosystems
Breininger, D. R., Stolen, E. D., Carter, G. M., Legare, S. A., Payne, W. V., Breininger, D. J., Lyon, J. E., Schumann, C., & Hunt, D. K. (2023) · Ecology and Evolution · doi:10.1002/ece3.9704 · Cited by 9
“This produced a distribution similar in shape to the Half- Cauchy distribution recommended by Gelman (2006), but with a much shorter tail.”
MethodFuture trajectory of respiratory infections following the COVID-19 pandemic in Hong Kong
Cheng, W., Zhou, H., Ye, Y., Chen, Y.-C., Jing, F., Cao, Z., Zeng, D. D., & Zhang, Q. (2023) · Chaos An Interdisciplinary Journal of Nonlinear Science · doi:10.1063/5.0123870 · Cited by 8
Comparing Methods of Data Synthesis
Oppe, M., Al, M., & Mölken, M. R. (2010) · PharmacoEconomics · doi:10.2165/11539870-000000000-00000 · Cited by 9
Interspecific relationships and environmentally driven catchabilities estimated from fisheries data
Hosack, G. R., Peters, G. W., & Ludsin, S. A. (2013) · Canadian Journal of Fisheries and Aquatic Sciences · doi:10.1139/cjfas-2013-0236 · Cited by 10
Golightly, A., Boys, R. J., Cameron, K., & von Zglinicki, T. (2012) · Journal of the Royal Statistical Society Series C (Applied Statistics) · doi:10.1111/j.1467-9876.2012.01045.x · Cited by 9
“The motivation for this approach is two-fold: this choice of prior leads to a Gibbs sampler that can easily be implemented and also is less influential on posterior inferences if the true state variances are close to zero than when using inverse Gamma priors (Gelman, 2006).”
Ball, S. J., Jacoby, P. A., & Zubrick, S. R. (2013) · International Journal of Environmental Research and Public Health · doi:10.3390/ijerph10072606 · Cited by 11
RYAN, L. M. (2009) · Epidemiology · doi:10.1097/ede.0b013e318198a5fb · Cited by 9
Collection, synthesis, and interpretation of evidence: a proof‐of‐concept study in COPD
Scala, L. D., Kerman, J., & Neuenschwander, B. E. (2013) · Statistics in Medicine · doi:10.1002/sim.5730 · Cited by 11
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.