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 2,601–2,700 of 4,089
Bayesian adjustment for measurement error in an offset variable in a Poisson regression model
Zhang, K., Liu, J., Liu, Y., et al. (2021) · Statistical Modelling · doi:10.1177/1471082x211008011 · Cited by 4
“Prior for 1/ √ τ: Following the suggestion in Gelman (2006), we employ uniform (0, M) prior for standard deviation.”
MethodComparing the performance of Bayesian and least-squares approaches for inverse kinematics problems
Pohl, A. J., Schofield, M., & Ferber, R. (2021) · Journal of Biomechanics · doi:10.1016/j.jbiomech.2021.110597 · Cited by 4
Baumann, U., Lodge, D. M., & Miescu, M. (2022) · International Journal of Finance & Economics · doi:10.1002/ijfe.2675 · Cited by 10
Modeling SF-6D Health Utilities: Is Bayesian Approach Appropriate?
Kharroubi, S. A. (2021) · International Journal of Environmental Research and Public Health · doi:10.3390/ijerph18168409 · Cited by 4
Bayesian Testing of Linear Versus Nonlinear Effects Using Gaussian Process Priors
Mulder, J. (2022) · The American Statistician · doi:10.1080/00031305.2022.2028675 · Cited by 6
“Further note that the half-Cauchy prior is becoming increasingly popular for scale parameters in Bayesian analyses (Gelman, 2006; Polson & Scott, 2012; Mulder & Pericchi, 2018).”
BackgroundBurger, D. A., Schall, R., & van der Merwe, S. (2021) · Pharmaceutical Research · doi:10.1007/s11095-021-03110-z · Cited by 4
Adjusting for misclassification of an exposure in an individual participant data meta‐analysis
de Jong, V. M. T., Campbell, H., Maxwell, L., Jaenisch, T., Gustafson, P., & Debray, T. P. A. (2022) · Research Synthesis Methods · doi:10.1002/jrsm.1606 · Cited by 4
“However, we did observe bias for estimators of the heterogeneity of the exposure-outcome relation (using all approaches), which is no surprise as estimating this parameter is notoriously difficult in meta-analyses of few studies.(29,42) We found that applying a misclassification model on the individual participant dat…”
MethodBayesian Model Selection for Longitudinal Count Data
Ariyo, O., Lesaffre, E. M. E. H., Verbeke, G. N., et al. (2021) · Sankhya B · doi:10.1007/s13571-021-00268-9 · Cited by 4
Chevalier, M., Tedesco, P. A., & Grenouillet, G. (2022) · PeerJ · doi:10.7717/peerj.12857 · Cited by 4
“For σ , σ 0 and σ > 0 + , we used half-Cauchy distributions ( Gelman, 2006 ).”
MethodPredicting Verbal Learning and Memory Assessments of Older Adults Using Bayesian Hierarchical Models
Ebrahim, E. A., & Cengiz, M. A. (2022) · Frontiers in Psychology · doi:10.3389/fpsyg.2022.855379 · Cited by 4
“Gelman (2006) suggested half-Cauchy prior with a mode at 0 and scale set to a considerable value by reasonably explaining the restrictive nature of half-Cauchy prior in providing enough information for the small numbers of groupings in the hierarchical nature of the data.”
MethodDaily appraisal of commercial real estate a new mixed frequency approach
Francke, M., & de Minne, A. V. (2022) · Real Estate Economics · doi:10.1111/1540-6229.12378 · Cited by 8
Pan, A., Song, X., & Huang, H. (2022) · Statistics in Medicine · doi:10.1002/sim.9531 · Cited by 4
Alam, F. M. A. (2022) · Austrian Journal of Statistics · doi:10.17713/ajs.v51i4.1331 · Cited by 4
“In fact, the halfCauchy distribution can be used in a Bayesian framework when a proper prior is necessary; see, for example, Gelman (2006); Polson and Scott (2012).”
MethodMarques, I., Kneib, T., & Klein, N. (2022) · Statistics and Computing · doi:10.1007/s11222-022-10136-9 · Cited by 9
“The fourth—and final—principle of userdefined scaling can also be applied to other prior structures, such as inverse gamma priors and proper uniform priors for standard deviations (Gelman 2005, 2006; Hodges 2019).”
BackgroundWasserman, A., Musella, A., Shapiro, M. A., & Shrager, J. (2022) · Artificial Intelligence in Medicine · doi:10.1016/j.artmed.2022.102450 · Cited by 4
Bayesian Design of Multi-Regional Clinical Trials with Time-to-Event Endpoints
Bean, N. W., Ibrahim, J. G., & Psioda, M. A. (2023) · Biometrics · doi:10.1111/biom.13820 · Cited by 4
Powell, J., Kalinowski, S. T., Taper, M. L., Rotella, J. J., Davis, C. S., & Garrott, R. A. (2023) · Entropy · doi:10.3390/e25030403 · Cited by 6
“half-Cauchy prior distributions was set to be four times larger than the observed standard deviations in the data [79].”
MethodMetcalfe, R. S., Swinton, P. A., Mackintosh, K. A., Berg, R. M. G., Shelley, J., SAYNOR, Z. L., Hudson, J., Duckers, J. M., Lewis, K. E., DAVIES, G. A., et al. (2023) · Medicine & Science in Sports & Exercise · doi:10.1249/mss.0000000000003207 · Cited by 4
Ma, S., Yu, K., Tang, M., Pan, J., Härdle, W. K., & Tian, M. (2023) · Statistics in Medicine · doi:10.1002/sim.9889 · Cited by 5
Kaplan, D. E., & Harra, K. (2024) · Large-scale Assessments in Education · doi:10.1186/s40536-023-00189-1 · Cited by 8
Han, D. C., Brownlow, J. D., Thompson, J., & Brooks, R. (2024) · Quality and Reliability Engineering International · doi:10.1002/qre.3606 · Cited by 4
Depaoli, S. A., & Scott, S. (2015) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705511.2015.1044653 · Cited by 4
RetractedMapping QTL for Multiple Traits
Xu, S. (2012) · Springer eBooks · doi:10.1007/978-0-387-70807-2_14 · Cited by 4
Fry, J. T., Hoegh, A. B., Leman, S. C., & Montesano, M. (2017) · Journal of Applied Statistics · doi:10.1080/02664763.2016.1276523 · Cited by 4
“of themodel, Gelman [6] found k = 1/2 to performwell in general settings.”
BackgroundApplication of SCM with Bayesian B-Spline to Spatio-Temporal Analysis of Hypertension in China
Ye, Z., Xu, L., Zhou, Z., et al. (2018) · International Journal of Environmental Research and Public Health · doi:10.3390/ijerph15010055 · Cited by 4
“Gelman acknowledged that BHM is a flexible tool for combining information and the partial pooling of inferences, providing a more objective method for inferences by estimating the parameters of prior distributions from data rather than requesting the data to be specified using subjective information [6].”
MethodWatson, S., Chen, Y., Bion, J., Aldridge, C., Girling, A. J., & Lilford, R. J. (2018) · BMJ Open · doi:10.1136/bmjopen-2016-015561 · Cited by 4
Analysis of Repeatability and Reproducibility Studies With Ordinal Measurements
Culp, S., Ryan, K. J., Chen, J., et al. (2018) · Technometrics · doi:10.1080/00401706.2018.1429317 · Cited by 4
“The necessary background on Bayesian data analysis is concisely summarized in online Appendix A. Gelman (2006) gave a relevant background on prior distributions and the language used in this article to describe the proposed prior distributions (e.g., improper vs. proper and informative vs. weakly…”
InfluentialMcCaffrey, D. F., Oliveri, M. E., & Holtzman, S. L. (2018) · ETS Research Report Series · doi:10.1002/ets2.12225 · Cited by 5
“We followed the approach described in Gelman (2006) in which the specified prior distribution for the square root of each variance component was a uniform distribution on a range chosen so that the specification of the prior distribution has little impact on the estimated variance components.3 We…”
MethodLiu, B., Wang, L., Nie, Y., & Cao, J. (2019) · Computational Statistics & Data Analysis · doi:10.1016/j.csda.2019.03.001 · Cited by 5
“Then the WAIC is given by (Gelman et al. 2014) WAIC = −2LPPD + 2pWAIC, where the penalty term, pWAIC, is used to correct the effective number of parameters. There are two different approaches to calculate this correction. Here, following the suggestion of Gelman et al. (2014), we use the variance version, pWAIC = n ∑…”
BackgroundMethodWU, J. M., Gupta, M., Hussein, A. I., & Gerstenfeld, L. C. (2020) · Journal of Applied Statistics · doi:10.1080/02664763.2020.1772733 · Cited by 4
A Bayesian Factor Model for Spatial Panel Data with a Separable Covariance Approach
Leorato, S., & Mezzetti, M. (2020) · Bayesian Analysis · doi:10.1214/20-ba1215 · Cited by 4
“The choice of the shape hyperparameters in inverse Wishart prior distributions has always been critical, especially when it is desired to achieve arbitrarily high noninformativity (Gelman (2006)).”
BackgroundDrug sensitivity prediction with normal inverse Gaussian shrinkage informed by external data
Münch, M., van de Wiel, M. A., Richardson, S. T., et al. (2020) · Biometrical Journal · doi:10.1002/bimj.201900371 · Cited by 4
“Gelman (2006), among others, advises against this practice.”
BackgroundCrash prediction for a French highway network with an XAI-informed Bayesian hierarchical model
Véran, T., Portier, P.-É., & Fouquet, F. (2020) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata50022.2020.9378376 · Cited by 4
“aGelman indicates in [18] that a half-normal with high standard deviation is a non-informative but proper prior for the variance parameter σ.”
BackgroundMeasuring the trend real interest rate in a data-rich environment
Fu, B. (2023) · Journal of Economic Dynamics and Control · doi:10.1016/j.jedc.2023.104606 · Cited by 4
Estimation of the Modified Weibull Additive Hazards Regression Model under Competing Risks
Rehman, H., Chandra, N., Emura, T., et al. (2023) · Symmetry · doi:10.3390/sym15020485 · Cited by 4
Combining multiple maps of line features to infer true position
Barber, J. J., & Prager, S. D. (2008) · Bayesian Analysis · doi:10.1214/08-ba325 · Cited by 4
Comparison of three-level cluster randomized trials using quantile dispersion graphs
Singh, S. P., Mukhopadhyay, S., & Roy, A. (2015) · Journal of Applied Statistics · doi:10.1080/02664763.2015.1010491 · Cited by 4
Embedding black-box regression techniques into hierarchical Bayesian models
Shaby, B. A., & Fink, D. (2011) · Journal of Statistical Computation and Simulation · doi:10.1080/00949655.2011.594052 · Cited by 6
“Gelman [19] recommends heavy-tailed priors like the Cauchy for variance components in hierarchical models, and we have found that these priors perform well, contributing almost no information, in random field models.”
MethodFormulation and Application of the Hierarchical Generalized Random-Situation Random-Weight MIRID
Hung, L.-F. (2011) · Multivariate Behavioral Research · doi:10.1080/00273171.2011.589274 · Cited by 4
“For models with more than two levels of random effects, Bayesian estimation methods can be very practical (Béguin & Glas, 2001; Congdon, 2003; De Boeck & Wilson, 2004; Gelman, 2006; Patz & Junker, 1999a; Wollack, Bolt, Cohen, & Lee, 2002).”
BackgroundMethodA reanalysis of a longitudinal scleroderma clinical trial using non-ignorable missingness models
Boscardin, W. J., Yan, X., & Wong, W. K. (2007) · Journal of Statistical Planning and Inference · doi:10.1016/j.jspi.2007.04.005 · Cited by 4
Testing for nonlinearity in mean and volatility for heteroskedastic models
Chen, C. W. S., Gerlach, R., & Tai, A. P. J. (2008) · Mathematics and Computers in Simulation · doi:10.1016/j.matcom.2008.01.044 · Cited by 4
“Such flat constrained prior choices are similar to those suggested by Gelman (2006) for positive parameters.”
ResultSensitivity analysis in Bayesian nonignorable selection model for binary responses
Choi, S. M., & Kim, D. H. (2014) · Journal of the Korean Data and Information Science Society · doi:10.7465/jkdi.2014.25.1.187 · Cited by 4
“Finally, we put p(τ) = 1 (1 + τ)2 , τ ≥ 0, which is called a shrinkage prior (see Gelman, 2006).”
Bayesian outlier analysis in binary regression
de Souza, A. D. P., & Migon, H. S. (2010) · Journal of Applied Statistics · doi:10.1080/02664760903031153 · Cited by 4
Thall, P., Liu, D. D., Berrak, S., & Wolff, J. (2011) · Statistics in Medicine · doi:10.1002/sim.4249 · Cited by 4
Estimating the prevalence of transmitted HIV drug resistance using pooled samples
Finucane, M. M., Rowley, C. F., Paciorek, C. J., et al. (2013) · Statistical Methods in Medical Research · doi:10.1177/0962280212473514 · Cited by 4
Rolfe, M. I., Mengersen, K. L., Vearncombe, K., Andrew, B., & Beadle, G. (2011) · Journal of the Royal Statistical Society Series C (Applied Statistics) · doi:10.1111/j.1467-9876.2011.00766.x · Cited by 3
“…context, the random effects for the intercept or level, slope and intercept– slope interaction can be estimated by using a range of prior distributions and include the inverse Wishart or Wishart, uniform and half-Cauchy distributions (Gelman and Hill, 2007; Gelman et al., 2004; Gelman, 2006).”
BackgroundMethodInfluentialTransforming Abandoned Farm Fields to Conifer Plantations Reduces Ruffed Grouse Density
Roy, C., Mazerolle, M. J., & Imbeau, L. (2015) · Journal of Fish and Wildlife Management · doi:10.3996/022015-jfwm-021 · Cited by 4
“We used normal priors with a mean of zero and half-Cauchy priors for the standard deviation (Gelman 2006).”
MethodSpatial Inequality in Wealth: A Bayesian Analysis of the Northeastern US in 1860 - Does Space Matter
Kasakoff, A. B., Lawson, A. B., Dasgupta, P., Feetham, S., & Dubois, M. (2013) · Spatial Demography · doi:10.1007/bf03354887 · Cited by 4
Bayesian assurance and sample size determination in the process validation life-cycle
Faya, P., Seaman, J. W., & Stamey, J. D. (2016) · Journal of Biopharmaceutical Statistics · doi:10.1080/10543406.2016.1148717 · Cited by 4
“Gelman (2006) has shown that, for two-level normal hierarchical models, using an IG(ε, ϵ) with small ε as a weakly informative prior on σ2B is problematic.”
BackgroundMethodInfluentialRandom partition models and complementary clustering of Anglo-Saxon place-names
Zanella, G. (2015) · The Annals of Applied Statistics · doi:10.1214/15-aoas884 · Cited by 4
“We use a flat uniform prior for σ , as, for example, it is recommended in Gelman (2006), Section 7.1: σ ∼ Unif(0, σmax).”
MethodInfluentialTransmission of monetary policy and exchange rate shocks under foreign currency lending
Skibińska, M. (2018) · Post-Communist Economies · doi:10.1080/14631377.2018.1442058 · Cited by 4
“For the prior for parameter 1 , Gelman (2006) and Jarociński (2010) recommend inverse Gamma distribution with small shape parameters or uniform uninformative prior.”
Bayesian inference in threshold stochastic frontier models
Tsionas, E. G., Tran, K. C., & Michaelides, P. G. (2017) · Empirical Economics · doi:10.1007/s00181-017-1364-9 · Cited by 5
Effects of prior distributions: An application to pipedwater demand
Hassan, A. R., & Pericchi, L. R. (2018) · Brazilian Journal of Probability and Statistics · doi:10.1214/16-bjps329 · Cited by 4
Bayesian Linear Mixed Models with Polygenic Effects
Zhao, J. H., Luan, J., & Congdon, P. (2018) · Journal of Statistical Software · doi:10.18637/jss.v085.i06 · Cited by 4
Ensemble coding of memory strength in recognition tests
Dubé, C., Tong, K., Westfall, H. A., & Bauer, E. (2019) · Memory & Cognition · doi:10.3758/s13421-019-00912-w · Cited by 4
“We chose an uninformative prior for the population mean standard deviation (Gelman, 2006) for Experiment 1 μσ1 and Experiment 2 μσ2: μσ1;μσ2∼Half −Gaussian 0; ffiffiffi 2 p : Similarly, the population standard deviations of the standard deviation for Experiment 1 σσ1 and Experiment 2 σσ2 each were…”
Improved estimation of SNP heritability using Bayesian multiple-phenotype models
Elhezzani, N. S. (2018) · European Journal of Human Genetics · doi:10.1038/s41431-018-0100-z · Cited by 5
Bayesian Autoregressive Frailty Models for Inference in Recurrent Events
Tallarita, M., Iorio, M. D., Guglielmi, A., & Malone‐Lee, J. (2019) · The International Journal of Biostatistics · doi:10.1515/ijb-2018-0088 · Cited by 5
“Gelman [30] as inverse-gamma distributions give numerical instability problems when hyperparameters are small.”
BackgroundWebb, J. A., Vietz, G. J., Windecker, S. M., et al. (2015) · Cited by 4
Gelman, A., & O’Rourke, K. (2014) · Springer eBooks · doi:10.1007/978-3-319-05467-4_10 · Cited by 4
Self-citationBayesian inference for quantiles of the log‐normal distribution
Gardini, A., Trivisano, C., & Fabrizi, E. (2020) · Biometrical Journal · doi:10.1002/bimj.201900386 · Cited by 4
“Similarly, the uniform prior over the range (0, 𝐴) for 𝜎 (Gelman, 2006) implies that 𝑝(𝜎2) ∝ 1 𝜎 𝟏(0,𝐴), which may be seen as an approximation to a 𝐺𝑎𝑚𝑚𝑎(1 2 , 𝜖) (where 𝜖 = (4𝐴2)−1) truncated at 𝐴2.”
BackgroundParameter Expansion and Efficient Inference
Lewandowski, A., Liu, C., & Wiel, S. V. (2010) · Cited by 4
Zahrieh, D., Golafshar, M. A., Patel, S. H., et al. (2021) · Preventing Chronic Disease · doi:10.5888/pcd18.200468 · Cited by 4
Bayesian inference for asymptomatic COVID‐19 infection rates
Cahoy, D., & Sedransk, J. (2022) · Statistics in Medicine · doi:10.1002/sim.9408 · Cited by 3
On the Practice of Bayesian Inference in Basic Economic Time Series Models Using Gibbs SamplingPreprint
Pooter, M. D. D., Segers, R., & van Dijk, H. K. (2006) · SSRN Electronic Journal · doi:10.2139/ssrn.929060 · Cited by 4
Commentary on ‘Multivariate meta‐analysis: potential and promise’
Harbord, R. M. (2011) · Statistics in Medicine · doi:10.1002/sim.4278 · Cited by 6
Simple, Scalable and Accurate Posterior Interval EstimationPreprint
Li, C., Srivastava, S., & Dunson, D. B. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1605.04029 · Cited by 4
“The model assumes that where gdP denotes the generalized double Pareto shrinkage prior of [3] and Half-t is chosen to be weakly-informative [10].”
Estimating the Mixed Logit Model by Maximum Simulated Likelihood and Hierarchical BayesPreprint
Akinc, D., & Vandebroek, M. L. (2017) · SSRN Electronic Journal · doi:10.2139/ssrn.3052293 · Cited by 5
“Halft priors for the standard deviations were recommended by Gelman(2006) as non-informative priors which do allow for small values.”
BackgroundMethodAchieving Parsimony in Bayesian VARs with the Horseshoe PriorPreprint
Follett, L., & Yu, C. L. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1709.07524 · Cited by 4
Dynamic Function-on-Scalars RegressionPreprint
Kowal, D. R. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1806.01460 · Cited by 4
“Since λfk > 0 corresponds to a prior precision parameter, we follow Gelman (2006) and Kowal et al. (2017a) and impose a uniform prior distribution on the corresponding standard deviation, λ −1/2 fk iid∼ Uniform(0, 104).”
MethodInfluentialLocally-adaptive Bayesian nonparametric inference for phylodynamicsPreprint
Faulkner, J. R., Magee, A. F., Shapiro, B., & Minin, V. N. (2018) · arXiv (Cornell University) · Cited by 4
Dynamic speech adaptation to unreliable cues during intonational processingPreprint
Roettger, T. B., & Franke, M. (2018) · OSF Preprints (OSF Preprints) · doi:10.31219/osf.io/v6exy · Cited by 5
“We used weakly informative Gaussian priors centered around zero with σ = 100 for all population-level regression coefficients (Gelman, 2006), as well as standard priors of the brms package for all other parameters.”
MethodBayesian Dynamic Fused LASSOPreprint
Irie, K. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1905.12275 · Cited by 4
“The horseshoe prior is also the scale mixture of normals; now the scale follows the half-Cauchy distribution (Gelman 2006) or, equivalently, the gamma mixture of gamma distributions.”
MethodThe Origins of Regional Integration: Untangling the Effect of Trade on Judicial CooperationPreprint
Lampach, N., & Dyèvre, A. (2017) · SSRN Electronic Journal · doi:10.2139/ssrn.3036797 · Cited by 3
Understanding rumination as a form of inner speech: probing the role of motor processesPreprint
Nalborczyk, L. (2019) · Thesis Commons (OSF Preprints) · doi:10.31237/osf.io/p6dct · Cited by 7
Smith, A. C., & Edwards, B. P. M. (2020) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2020.03.26.010215 · Cited by 10
A Robust Bayesian Copas Selection Model for Quantifying and Correcting Publication BiasPreprint
Bai, R., Lin, L., Boland, M. R., & Chen, Y. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2005.02930 · Cited by 11
Zeng, S., Li, F., & Ding, P. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2005.09130 · Cited by 4
Daily Appraisal of Commercial Real Estate A New Mixed Frequency ApproachPreprint
Francke, M., & de Minne, A. V. (2021) · SSRN Electronic Journal · doi:10.2139/ssrn.3789205 · Cited by 4
The Energetic Cost of Anthropogenic Disturbance on the Southern Sea Otter (Enhydra lutris nereis)
Barrett, H. (2019) · doi:10.31979/etd.62cf-qybg · Cited by 4
Making more with forest inventory data: Toward a scalable, dynamical model of forest changePreprint
Itter, M. S., & Finley, A. O. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.07.22.604669 · Cited by 4
“Among species standard deviation parameters ( τ ℓ ) are assigned half-Cauchy prior distributions (Gelman, 2006).”
de Hollander, G., Grueschow, M., Hennel, F., et al. (2024) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2024.08.23.609296 · Cited by 4
Pettini, L., Bogler, C., Matić, K., Görgen, K., Doeller, C. F., & Haynes, J. (2025) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2025.12.11.693774 · Cited by 6
Huh, Y. K., Kim, M., Olivas, K., Allen, T. R., & Liu, K. (2024) · Journal of Quality Technology · doi:10.1080/00224065.2024.2369674 · Cited by 3
“Here, we use informative prior distributions (Gelman 2006) inspired by domain knowledge to overcome the imbalanced data challenge mentioned in Section 1.2.”
InfluentialFang, G., Pan, R., Wang, Y., et al. (2024) · IEEE Transactions on Reliability · doi:10.1109/tr.2024.3471796 · Cited by 3
“For the variance components σ 20 , σ 21 , and σ 2 , the half-Cauchy distribution that ensures positivity is recommended instead of the inverse-gamma and uniform distributions according to Gel-man [38].”
Huber, M., Müller, T., & Schmitt, R. H. (2025) · Measurement Sensors · doi:10.1016/j.measen.2024.101608 · Cited by 3
Park, J., Sim, K., Yang, D., Kim, Y., Hashizume, M., & Chung, Y. (2025) · Environmental and Ecological Statistics · doi:10.1007/s10651-025-00655-9 · Cited by 5
Data reconstruction from machine learning models via inverse estimation and Bayesian inference
Hartoyo, A., Ciupek, D., Malawski, M., & Crimi, A. (2025) · Scientific Reports · doi:10.1038/s41598-025-96215-z · Cited by 3
Bhardwaj, U., Teixeira, A. P., & Soares, C. G. (2025) · Journal of Loss Prevention in the Process Industries · doi:10.1016/j.jlp.2025.105666 · Cited by 3
Matavel, C., Meyer‐Aurich, A., & Piepho, H. (2025) · Agronomy Journal · doi:10.1002/agj2.70087 · Cited by 3
Foraging in proximity to humans can shape social centrality in wild dolphins
Bankhead, K., McHugh, K., Wells, R. S., & Cantor, M. (2025) · Behavioral Ecology and Sociobiology · doi:10.1007/s00265-025-03641-0 · Cited by 5
Monitoring Pacific walrus coastal haulouts by satellite to estimate herd abundance and distribution
Fischbach, A. S., Taylor, R. L., & Douglas, D. C. (2025) · Wildlife Society Bulletin · doi:10.1002/wsb.1614 · Cited by 3
Depaoli, S. A., Qiu, M., Liu, H., & Jauregui, M. (2025) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705511.2025.2563187 · Cited by 3
“An inverse gamma prior InvGamma ð 1, 1 Þ was selected for the variances of the measurement errors / 2 t ; which is considered relatively informative and has been recommended by Gelman (2006) to avoid improper posteri-ors.”
Free odor identification engages domain-general cognitive abilities in old adults
Hörberg, T., Olofsson, J., Raj, R., Laukka, E. J., & Larsson, M. (2025) · Chemical Senses · doi:10.1093/chemse/bjaf049 · Cited by 3
“All continuous variables were standardized by subtracting their sample mean and dividing by two standard deviations (following Gelman 2006: 57).”
Amdi, C. H., Spence, A., Helms, E. R., & McGuigan, M. R. (2025) · The Journal of Strength and Conditioning Research · doi:10.1519/jsc.0000000000005261 · Cited by 5
Yang, M., Pan, S., Wang, Q., Jiang, Y., Li, X., Fan, X., Zhan, D., & Wang, D. (2025) · BMC Pediatrics · doi:10.1186/s12887-025-06433-7 · Cited by 3
Paired comparison models with strength-dependent ties and order effects
Glickman, M. E. (2025) · Statistical Modelling · doi:10.1177/1471082x251400474 · Cited by 3
“The variance parameters σ 2 and σ were assumed to follow half-Cauchy distributions with location parameter 0 and scale parameter 5, following recommendations for weakly informative priors on scale parameters in hierarchical models (Gelman, 2006).”
Bayesian hierarchical cognitive modeling with the EMC2 package
Stevenson, N., Donzallaz, M. C., Innes, R. J., Forstmann, B. U., Matzke, D., & Heathcote, A. (2026) · Behavior Research Methods · doi:10.3758/s13428-025-02869-y · Cited by 9
“:3) that do not include covariances, the prior on the diagonal entries ( σ 21 ... p ) is a mixture of inverse-Gamma distributions that gives rise to a half-t prior distribution on the standard deviations (Wand, Ormerod, Padoan, & Frühwirth, 2011; Gelman, 2006).”
Art knowledge training shapes understanding, inspires creativity and stimulates thinking
Bara, I., Ramsey, R., & Cross, E. S. (2026) · Royal Society Open Science · doi:10.1098/rsos.250805 · Cited by 3
“Also, by using weakly informative priors, we allow for the possibility of large effects, should they exist in the data [70–73].”
Rast, P., & Clayson, P. E. (2026) · British Journal of Mathematical and Statistical Psychology · doi:10.1111/bmsp.70026 · Cited by 6
Shahbazi, M. A., Baheri, A., & Azadeh‐Fard, N. (2026) · Scientific Reports · doi:10.1038/s41598-026-37450-w · Cited by 8
“…Modeling and Uncertainty Quantification Bayesian methods provide principled uncertainty quantification through posterior distribution estimation, with Bayesian hierarchical models specifically designed to capture multi-level uncertainty through partial pooling across nested structures 26,45 .”
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.