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 3,201–3,300 of 4,089
Asymmetrical triadic relationship based on the structural difficulty
Kosugi, K. E. (2017) · Behaviormetrika · doi:10.1007/s41237-017-0033-9 · Cited by 1
Regan, T., Paul, C. L., Ishiguchi, P., D’Este, C. A., Koller, C. E., Fakes, K., Noble, N., Oldmeadow, C. J., Bisquera, A., & Eades, S. (2017) · International Journal of Environmental Research and Public Health · doi:10.3390/ijerph14101236 · Cited by 1
“Unlike frequentist methods that are generally poor at estimating between-group variances when there are a low number of groups [17], the Bayesian approach has been shown to produce reasonable estimates [18].”
MethodModel fitting and evaluation in climate reconstruction of tree-ring data
Schofield, M., & Barker, R. (2017) · Dendrochronologia · doi:10.1016/j.dendro.2017.10.004 · Cited by 1
Outlier detection in contamination control
Warrick, S. P., & Weintraub, J. (2018) · doi:10.1117/12.2297379 · Cited by 1
Accounting for measurement error in analyzing data
Hamada, M. S., & Weaver, B. (2018) · Quality Engineering · doi:10.1080/08982112.2018.1518530 · Cited by 1
密度推定に基づいたタヌキに対する外来哺乳類(アライグマ・ハクビシン)の影響
栗山, 武., 小井土, 美., 長田, 穣., 正彦, 浅., Yokomizo, H., & 宮下, 直. (2018) · 保全生態学研究 · doi:10.18960/hozen.23.1_9 · Cited by 4
Li, L., Lee, J., Sutton, S. K., et al. (2019) · Statistical Modelling · doi:10.1177/1471082x18821489 · Cited by 1
“For the covariance matrix , we adopt the hierarchical half-t prior in Huang and Wand (2013), which is a multivariate extension of the half-t prior in Gelman (2006).”
MethodLuo, Y., & Liang, X. (2019) · Measurement Interdisciplinary Research and Perspectives · doi:10.1080/15366367.2018.1533783 · Cited by 1
“For model converge check, we applied the convergence statistic proposed by Gelman and Rubin (1992) that computes the potential scale reduction factor (PSRF).”
MethodWAND, M. P. (2016) · Journal of the American Statistical Association · doi:10.1080/01621459.2016.1197833 · Cited by 41
“As explained in Gelman (2006), approximate noninformativity of σ u can be achieved via Uniform distribution and Half t distribution priors.”
Rubio, F. J. (2014) · Statistics & Probability Letters · doi:10.1016/j.spl.2014.09.023 · Cited by 1
“text has also foster some discussions. In particular, the inverse gamma distribution, which appears in the structure (4), has been shown to be influential on the shape 5 of the posterior distribution (Gelman, 2006). This is an undesirable property if one is interested on using weakly informative priors (which is usuall…”
BackgroundMasaoud, E., & Stryhn, H. (2020) · Journal of Statistical Research · doi:10.47302/jsr.2020540101 · Cited by 1
“The uniform distribution for inverse variances, or precisions (τ ∼ uniform(0,100)) was used (Lambert et al., 2005; Gelman , 2006).”
MethodA comparative study on high-dimensional bayesian regression with binary predictors
Slanzi, D., Mameli, V., & Brown, P. J. (2021) · Communications in Statistics - Simulation and Computation · doi:10.1080/03610918.2021.1894337 · Cited by 1
“This prior for the regression coefficients b j can be obtained by assuming zero-mean independent half-Cauchy priors for all the parameters s 2 j , where C þ ð 0, 1 Þ denotes a standard half-Cauchy distribution (Gelman 2006).”
MethodDependent Dirichlet Processes for Analysis of a Generalized Shared Frailty Model
Zhong, C., Ma, Z., Shen, J., et al. (2022) · IntechOpen eBooks · doi:10.5772/intechopen.101502 · Cited by 1
“Particularly, inspired by [27] and [28], we use the half-Cauchy distribution as the non-informative prior for the variance parameter instead of the inverse Gamma prior.”
MethodJoint analysis of preferences and drop out data in discrete choice experiments
Maaya, L., Meulders, M., & Vandebroek, M. L. (2021) · Journal of Choice Modelling · doi:10.1016/j.jocm.2021.100308 · Cited by 1
Accounting for post‐randomization variables in meta‐analysis: A joint meta‐regression approach
Lian, Q., Zhang, J., Hodges, J. S., Chen, Y., & Chu, H. (2021) · Biometrics · doi:10.1111/biom.13573 · Cited by 2
Kucharska‐Newton, A., Loop, M. S., Bulló, M., Moore, C., Haas, S. W., Wagenknecht, L. E., Whitsel, E. A., & HEISS, G. M. (2021) · International Journal of Cardiology · doi:10.1016/j.ijcard.2021.12.022 · Cited by 2
A flexible Bayesian framework for individualized inference via adaptive borrowing
Ji, Z., & Wolfson, J. A. (2021) · Biostatistics · doi:10.1093/biostatistics/kxab051 · Cited by 3
Modeling Biomarker- Informed Adaptive Design
Wang, J. (2017) · Journal of Bioanalysis & Biostatistics · doi:10.13188/2641-8681.1000005 · Cited by 1
Wakefield, J. (2012) · Springer series in statistics · doi:10.1007/978-1-4419-0925-1_2 · Cited by 2
Ramakrishnan, V. M., Niemann, T., Maletzki, P., Guenther, E., Bujaroska, T., Labulo, O., Li, Z., Slieker, J. C., Huch, R., Lehmann, K. H., et al. (2022) · Diagnostics · doi:10.3390/diagnostics12040785 · Cited by 1
“For parameter estimation, a full Bayesian model was deployed [9].”
MethodThink like a Bayesian and avoid pitfalls from our frequentist past
Doll, J. C., & Feiner, Z. S. (2022) · Ideas in Ecology and Evolution · doi:10.24908/iee.2022.15.1.c · Cited by 2
Crozet, M., Rivier, C., Lourenço, V., et al. (2022) · Talanta · doi:10.1016/j.talanta.2022.123394 · Cited by 1
Bayesian learning of Covid-19 vaccine safety while incorporating adverse events ontology
Zhao, B., Zhong, Y., Kang, J., & Zhao, L. (2023) · The Annals of Applied Statistics · doi:10.1214/23-aoas1743 · Cited by 2
Fahrmeir, L., Kneib, T., Lang, S., & Marx, B. D. (2013) · Springer eBooks · doi:10.1007/978-3-642-34333-9_7 · Cited by 1
Posterior propriety of an objective prior for generalized hierarchical normal linear models
Lin, C., Sun, D., & Song, C. (2022) · Statistical Theory and Related Fields · doi:10.1080/24754269.2021.1978206 · Cited by 1
Assessing dependence between frequency and severity through shared random effects
Becker, D. G., Woolford, D. G., & Dean, C. B. (2022) · PLoS ONE · doi:10.1371/journal.pone.0271904 · Cited by 1
“The prior distribution for the variance parameters of bi and Xijr were set as uniform on the interval (0, 20), based on the recommendations in [41].”
MethodIncorporating Variation and Quality of the Underlying Effects in Meta-Analysis
Fu, J., & Lin, J. (2022) · Journal of Systems Science and Complexity · doi:10.1007/s11424-022-1429-5 · Cited by 1
“Here, we pick out vague priors for μ and τ , that’s to say, Normal(0, 100) for μ and Uniform(0, 100) for τ ([32]).”
MethodOtsu, C., Iijima, H., & Nagaike, T. (2022) · Journal of Forest Research · doi:10.1080/13416979.2022.2133773 · Cited by 1
Bayesian estimation of Gegenbauer processes
Hunt, R. L., Peiris, S., & Weber, N. C. (2022) · Journal of Statistical Computation and Simulation · doi:10.1080/00949655.2022.2138883 · Cited by 1
A Bayesian method for multinomial probit model
Koo, D., Kim, C., & Lee, K. (2022) · Journal of the Korean Statistical Society · doi:10.1007/s42952-022-00200-5 · Cited by 1
“In particular, for high-dimensional problems, the usual inverse-Wishart prior is not sufficiently “non-informative” to deal with each element in the covariance matrix flexibly (Barnard et al., 2000; Gelman, 2006).”
BackgroundHow to Improve MCMC Convergence
Matsuura, K. (2022) · Springer eBooks · doi:10.1007/978-981-19-4755-1_9 · Cited by 1
A dynamic causal modeling of the second outbreak of COVID-19 in Italy
Bilancia, M., Vitale, D., Manca, F., Perchinunno, P., & Santacroce, L. (2023) · AStA Advances in Statistical Analysis · doi:10.1007/s10182-023-00469-9 · Cited by 1
“The specification of the prior distributions over the model parameters is as follows: (2) (5) This joint prior specification is a generic weakly informative choice that attempts to introduce weak prior assumptions about the true location of the unknown parameters into our analysis (Gelman 2006).”
Nomura, S., Ohigashi, T., & Sawamoto, R. (2022) · Japanese Journal of Biometrics · doi:10.5691/jjb.43.63 · Cited by 3
Witte, D. D., Abad, A. A., Molenberghs, G., et al. (2023) · Spatial and Spatio-temporal Epidemiology · doi:10.1016/j.sste.2023.100588 · Cited by 1
Bayesian Conway-Maxwell-Poisson (CMP) regression for longitudinal count data
Alam, M., Gwon, Y., & Meza, J. L. (2023) · Communications for Statistical Applications and Methods · doi:10.29220/csam.2023.30.3.291 · Cited by 1
Shergina, E., Richter, K. P., Daley, C. M., Faseru, B., Choi, W. S., & Gajewski, B. J. (2023) · Journal of Biopharmaceutical Statistics · doi:10.1080/10543406.2023.2233598 · Cited by 1
“Gelman suggests considering inverse gamma, uniform or half-Cauchy as a hyperprior distribution for variance of the normally distributed parameter of interest (Gelman 2006).”
BackgroundBayesian adaptive design for covariate‐adaptive historical control information borrowing
Jin, H., Kim, M., Scheffler, A., & Jiang, F. (2023) · Statistics in Medicine · doi:10.1002/sim.9913 · Cited by 4
Bayesian Latent Gaussian Models
Hrafnkelsson, B., & Bakka, H. C. (2023) · Springer eBooks · doi:10.1007/978-3-031-39791-2_1 · Cited by 5
ANOMALY DETECTION IN OFFSHORE WIND TURBINE STRUCTURES USING HIERARCHICAL BAYESIAN MODELLING
SMITH, S. M., HUGHE, A. J., DARDENO, T. A., et al. (2023) · International Workshop on Structural Health Monitoring (IWSHM) · doi:10.12783/shm2023/37051 · Cited by 1
“Gelman et al. [17] recommends the use of HalfCauchy distributions for population variances, as their heavy tails bring a robustness against outliers to the model, as well as efficiency during the inference and sampling process.”
MethodSu, L., Chen, X., Zhang, J., & Yan, F. (2023) · Journal of Biopharmaceutical Statistics · doi:10.1080/10543406.2023.2292211 · Cited by 1
“Instead of using the classic inverse-gamma prior, which has undesirable properties in the presence of heterogeneity as extensively discussed by Gelman (2006), we assign a more robust half-Cauchy prior distribution to the standard deviation parameter σ m .”
Saitoh, T., & Cohen, J. E. (2024) · Population Ecology · doi:10.1002/1438-390x.12176 · Cited by 1
“We tested the validity of a gamma distribution (0.001, 0.001), which we used previously, for the non-informative prior distribution for σ j , because Gelman (2006) suggested that a gamma distribution might cause a biased estimate for a very small σ j and recommended instead a uniform distribution…”
InfluentialPeterson, E. N., Guranich, G., Cresswell, J. A., & Alkema, L. (2024) · Statistics and Public Policy · doi:10.1080/2330443x.2023.2286313 · Cited by 1
“The first-order differences, ξ c , t are modeled with a zero-mean bivariate normal given by For information on prior assumptions, refer to (Peterson et al. 2022; Gelman 2006).”
Handling dependent samples in meta-analytic structural equation models: A Wishart-based approach
Uanhoro, J. (2024) · Behavior Research Methods · doi:10.3758/s13428-024-02340-4 · Cited by 1
“(cid:12) Both σ λ and residual standard deviations ( diag ( (cid:6) ) ) have half-t priors (Gelman, 2006) creating a weakly regularizing effect (Lemoine, 2019) given that true the total variable variances are 1.”
Chi, X., Yuan, Y., Yu, Z., & Lin, R. (2024) · Biometrical Journal · doi:10.1002/bimj.202300122 · Cited by 3
Luan, H., Ransome, Y., Dean, L. T., Nassau, T., & Brady, K. A. (2024) · Geographical Analysis · doi:10.1111/gean.12391 · Cited by 1
A Bayesian nonparametric meta‐analysis model for estimating the reference interval
Cao, W., Chu, H., Hanson, T., et al. (2024) · Statistics in Medicine · doi:10.1002/sim.10001 · Cited by 1
Efficient sparsity-promoting MAP estimation for Bayesian linear inverse problems
Lindbloom, J., Glaubitz, J., & Gelb, A. (2024) · Inverse Problems · doi:10.1088/1361-6420/ada17f · Cited by 7
“One variance hyper-prior that is frequently used in the literature (e.g. see [5, 9, 25, 26, 29, 49]) is an ‘uninformative’ inverse gamma distribution 4 , which corresponds to the choice of hyper-prior parameters ˜ r = − 1 , ˜ β = 1 , ˜ ϑ ≈ 0.”
Bayesian inference for amplitude distribution with application to radar clutter
Teimouri, M., Hoseini, S. M., & Greco, M. (2024) · Digital Signal Processing · doi:10.1016/j.dsp.2024.104443 · Cited by 1
Gaussian random fields on the product of spheres: Theory and applications
Alegría, A., Cleanthous, G., Georgiadis, A. G., et al. (2024) · Electronic Journal of Statistics · doi:10.1214/24-ejs2231 · Cited by 1
Hupf, B., Yang, Y., Gryder, R. W., Bunn, V., & Lin, J. (2024) · Journal of Biopharmaceutical Statistics · doi:10.1080/10543406.2024.2330206 · Cited by 1
“Other choices of priors when F is the binomial distribution are discussed in Spiegelhalter et al. (2004), Gelman (2006), and Polson and Scott (2012).”
Chen, I., Wu, Z., Harlow, S. D., Karvonen-Gutierrez, C. A., Hood, M. M., & Elliott, M. R. (2024) · The Annals of Applied Statistics · doi:10.1214/23-aoas1852 · Cited by 3
“In this setting, inferences using the inverse-Gamma distribution are extremely sensitive to the choice of hyperparameter values [Gelman, 2006, p. 524], which makes the inverse-Gamma prior “not at all uninformative”.”
MethodExploring an EM-algorithm for banded regression in computational neuroscience
Fuglsang, S. A., Madsen, K. H., Puonti, O., Siebner, H. R., & Hjortkjær, J. (2024) · Imaging Neuroscience · doi:10.1162/imag_a_00155 · Cited by 3
Bayesian Hierarchical Models of Basket Trials in the Context of Economic Evaluation
CADTH (2024) · Canadian Journal of Health Technologies · doi:10.51731/cjht.2024.907 · Cited by 1
“37 The degrees of freedom increase the certainty of the half-t distribution, but the scale increases the uncertainty.”
BackgroundKiffney, P. M., Anderson, J. H., Liermann, M. C., Jones, E. L., Pess, G. R., & Kretschmer, F. (2024) · Restoration Ecology · doi:10.1111/rec.14209 · Cited by 1
“We used a half Cauchy prior with variance = 1 for the standard deviation parameters, σ (Gelman 2006).”
Shan, N., & Xu, P. (2024) · Psychometrika · doi:10.1007/s11336-024-09998-x · Cited by 2
Wake up and search for coffee: Considering the circadian rhythm of consumers on online marketplaces
Liang, Z., Angelopoulos, S. P., Zuo, M., & Ou, C. X. (2024) · Journal of Business Research · doi:10.1016/j.jbusres.2024.114901 · Cited by 2
Early‐life diet composition affects phenotypic variation of correlated animal personality traits
Serrano‐Davies, E., de Miguel, A., Sepers, B., et al. (2024) · Ecology and Evolution · doi:10.1002/ece3.11567 · Cited by 1
“We used a parameter expanded and weakly informative prior for our model ( V = 3, nu = 1.002; Gelman, 2006; Houslay & Wilson, 2017) and ran 750,000 iterations with a burn-in of 50,000 and a thinning interval of 175 (effective sample size = 4000).”
MethodInfluentialThe Diffusion of Exclusion: Medieval Expulsions of Jews
Doten-Snitker, K. (2024) · Comparative Political Studies · doi:10.1177/00104140241269956 · Cited by 1
“I use the Student ’ s half-t distribution (df = 4, scale = 1) for weakly informative prior probability distributions for the intercepts (Gelman, 2006).”
Upton, M., Parnell, A. C., Kemp, A., Ashe, E., McCarthy, G., & Cahill, N. (2024) · Journal of the Royal Statistical Society Series C (Applied Statistics) · doi:10.1093/jrsssc/qlae044 · Cited by 5
Generalized sinh-Gaussian random fields
Homayouni, M., & Khaledi, . J. (2024) · Communication in Statistics- Theory and Methods · doi:10.1080/03610926.2024.2415379 · Cited by 1
Yu, M., Wu, Z., Hicken, M. T., et al. (2024) · Statistics in Medicine · doi:10.1002/sim.10281 · Cited by 1
“To allow sharing of information across individuals for long-term variability, we propose a common prior distribution for with hyper-parameter priors (Gelman, 2006) A general form for the prior distribution for the subject-level correlation matrix R i is difficult to specify for n i > 2 , given the…”
BackgroundFast Computation of Randomly Walking Volatility with Chained Gamma Distributions
Zhang, D., & Zhou, Y. (2024) · Computational Economics · doi:10.1007/s10614-024-10777-0 · Cited by 1
“Subsequently, Gelman (2004) proposed switching to an inverse gamma distribution (or equivalently, modeling all returns’ precision (i.e., 1/sigma) as gamma) to preserve the posterior’s form during Bayesian inference.”
A unit-level one-inflated beta model for small area prediction of seat-belt use rates
Zhou, Z., & Berg, E. J. (2024) · Journal of Applied Statistics · doi:10.1080/02664763.2024.2426016 · Cited by 1
“The reference [9] recommends the half-Cauchy prior for positive-valued parameters, as it is part of the conditionally-conjugate, folded-noncentral-t distribution family.”
Alt, E. M., Qu, Y., Damone, E., Liu, J., Wang, C., & Ibrahim, J. G. (2025) · Statistics in Medicine · doi:10.1002/sim.70021 · Cited by 2
Chimpanzees adapt their exploration to key properties of the environment
Haux, L. M., Engelmann, J. M., Herrmann, E., & Hertwig, R. (2025) · Nature Communications · doi:10.1038/s41467-025-57022-2 · Cited by 1
“We speci fi ed weakly informative normal priors with mean 0 and standard deviation 2 on all population-level effects 34 .”
Bayesian Growth Curve Modeling with Measurement Error in Time
Zhang, L., Qu, W., & Zhang, Z. J. (2025) · Multivariate Behavioral Research · doi:10.1080/00273171.2025.2473937 · Cited by 1
Sarkar, P., Khare, K., & Ghosh, M. (2025) · Bernoulli · doi:10.3150/24-bej1810 · Cited by 1
Chen, S., Li, Z., & Lü, X. (2025) · Behavioral Ecology · doi:10.1093/beheco/araf031 · Cited by 1
Ecological inference for relative risks, with application to infrequent mental health events
Congdon, P. D. (2025) · Journal of Geographical Systems · doi:10.1007/s10109-025-00457-4 · Cited by 1
Beyond conditional averages: Estimating the individual causal effect distribution
Post, R., & van den Heuvel, E. R. (2025) · Journal of Causal Inference · doi:10.1515/jci-2024-0007 · Cited by 7
Khalaj, S., Ba, H. T., Cholette, M. E., et al. (2025) · International Journal of Rail Transportation · doi:10.1080/23248378.2025.2508803 · Cited by 1
A Geometric Perspective on Bayesian and Generalized Fiducial Inference
Liu, Y., Hannig, J., & Murph, A. C. (2025) · Statistical Science · doi:10.1214/24-sts928 · Cited by 2
Zong-yuan, Z., & Luo, Q. (2025) · Applied Economics · doi:10.1080/00036846.2025.2516161 · Cited by 2
“The classical DLM model typically assumes that the errors follow a Gaussian distribution, and when Bayesian methods are used to estimate the unknown parameters, the parameter priors are usually chosen to be the inverse gamma distribution (Andrew 2006).”
Sparse 2-stage Bayesian meta-analysis for individualized treatments
Shen, J., Moodie, E. E. M., & Golchi, S. (2025) · Biometrics · doi:10.1093/biomtc/ujaf082 · Cited by 2
“Popula r prior choices include a normal prior with l arge v ari ance for the mean parameter ψ t and a half-Cauchy prior for the v ari anc e c omponent parameter σ ψ t (Gelman, 2006 ; Gelman et al., 2013 ).”
Understanding the COVID-19 pandemic through bayesian spatio-temporal modeling of several outcomes
Rodriguez-Idiazabal, L., Martínez‐Beneito, M. A., Quintana, J. M., García-Asensio, J., Legarreta, M. J., Larrea, N., & Barrio, I. (2025) · Spatial and Spatio-temporal Epidemiology · doi:10.1016/j.sste.2025.100737 · Cited by 4
Massa, F., Scavino, M., & Terrera, G. M. (2025) · Stats · doi:10.3390/stats8030074 · Cited by 1
“Finally, as suggested by [22,23], we assigned half-Cauchy distributions as a weakly informative prior for the parameters associated with the error and random effects variances.”
King, J. G., Pick, J. L., & Hadfield, J. D. (2025) · Methods in Ecology and Evolution · doi:10.1111/2041-210x.70087 · Cited by 3
“…problems, including a priori dependencies between variances and correlations (Tokuda et al., 2011), marginal distributions for the variances (inverse-Gamma) with high density around zero (Gelman, 2006) and a single degree of freedom controlling all variances and correlations (Gelman et al., 2013 ).”
Flexible Bayesian Estimation of Elliptical Copulas
Liang, P., & Rosen, O. (2025) · Journal of Computational and Graphical Statistics · doi:10.1080/10618600.2025.2573152 · Cited by 1
“The parameter τ is assumed to follow a Half-t distribution (Gelman 2006), with PDF where the hyperparameters n τ and A are assumed known.”
Filiberti, E., Roth, A. M., Thogmartin, W. E., et al. (2025) · The Auk · doi:10.1093/ornithology/ukaf049 · Cited by 1
Microarray Differential Expression Analysis
Xu, S. (2012) · Springer eBooks · doi:10.1007/978-0-387-70807-2_17 · Cited by 1
Fitting More Complex Bayesian Models: Markov Chain Monte Carlo
Cowles, M. K. (2012) · Springer texts in statistics · doi:10.1007/978-1-4614-5696-4_8 · Cited by 1
Regression and Hierarchical Regression Models
Cowles, M. K. (2012) · Springer texts in statistics · doi:10.1007/978-1-4614-5696-4_10 · Cited by 1
Generalization of rejection and acceptance in social networks
Zhang, Y., & Hackel, L. M. (2025) · Journal of Experimental Social Psychology · doi:10.1016/j.jesp.2025.104834 · Cited by 2
Nonparametric Regression with Multiple Predictors
Wakefield, J. (2012) · Springer series in statistics · doi:10.1007/978-1-4419-0925-1_12 · Cited by 1
Hypothesis Testing and Variable Selection
Wakefield, J. (2012) · Springer series in statistics · doi:10.1007/978-1-4419-0925-1_4 · Cited by 1
Geographical Disparities in Mortality Rates: Spatial Data Mining and Bayesian Hierarchical Modeling
Bilancia, M., Graziano, G., & Demarinis, G. (2012) · Contributions to statistics · doi:10.1007/978-88-470-2751-0_1 · Cited by 1
Bayesian Hierarchical Modeling of Growth via Gompertz Model: An Application in Poultry
Karaman, E., Kaya, E., Narinç, D., et al. (2013) · Springer proceedings in mathematics & statistics · doi:10.1007/978-3-319-02084-6_21 · Cited by 1
Hierarchical Clustering of Microarray Data
Xu, S. (2012) · Springer eBooks · doi:10.1007/978-0-387-70807-2_18 · Cited by 1
Bayesian random effects selection in mixed accelerated failure time model for interval-censored data
Harun, N., & Cai, B. (2013) · Statistics in Medicine · doi:10.1002/sim.6002 · Cited by 1
“Kinney and Dunson [7] extended this approach by using a new default prior structure based on a multivariate generalization of Gelman’s [8] folded-t prior.”
BackgroundMethodInfluentialHuang, Y., Yan, C., Wu, H., et al. (2014) · Statistics in Biopharmaceutical Research · doi:10.1080/19466315.2014.916627 · Cited by 1
“…(ν1 > 2), ν2 ∼ exp(0.5)I (ν2 > 2), (12) where 1=diag(0.01, 0.01, 0.01, 0.01), 2=diag(0.01, 0.01, 0.01); the mutually independent uniform (U), Normal (N), Exponential (exp), and inverse Wishart (IW) prior distributions are chosen to facilitate computations (Gelman 2006; Carlin and Louis 2009).”
BackgroundInfluentialWakefield, J. (2012) · Springer series in statistics · doi:10.1007/978-1-4419-0925-1_7 · Cited by 1
Akıner, M. E., & Akkoyunlu, A. (2016) · Archives of Environmental Protection · doi:10.1515/aep-2016-0019 · Cited by 1
Hierarchical Bayes Small Area Estimation under a Unit Level Model with Applications in Agriculture
Nazir, N., Mir, S., & Bhat, M. A. (2016) · Pakistan Journal of Statistics and Operation Research · doi:10.18187/pjsor.v12i3.1308 · Cited by 1
“Any reasonable noninformative prior usually works well for 2e (Gelman, 2006).”
BackgroundStatistical Methods for the Detection and Analysis of Radioactive Sources
Klumpp, J. A., Brandl, A., Johnson, T., et al. (2007) · Digital Collections of Colorado (Colorado State University) · doi:10.13140/2.1.3397.2003 · Cited by 1
Bayesian spatial analysis of obesity proportion data
Choi, J. (2016) · Journal of the Korean Data and Information Science Society · doi:10.7465/jkdi.2016.27.5.1203 · Cited by 1
Sheng, Y., & LeSage, J. P. (2016) · Emerald Group Publishing Limited eBooks · doi:10.1108/s0731-905320160000037019 · Cited by 1
A hierarchical Bayesian model for binary data incorporating selection bias
Choi, S., Nandram, B., & Kim, D. (2016) · Communications in Statistics - Simulation and Computation · doi:10.1080/03610918.2015.1130837 · Cited by 1
“This latter prior is used to avoid the difficulties associated with improper priors of the form p(τ ) ∝ 1/τ (e.g., see Gelman, 2006).”
MethodReviews of Books and Teaching Materials
Ramezan, R. (2017) · The American Statistician · doi:10.1080/00031305.2017.1271242 · Cited by 1
Ziniti, B. L. (2016) · University of New Hampshire Scholars Repository (University of New Hampshire at Manchester) · Cited by 1
Abbruzzini, T. F., Brandani, C. B., Toledo, F. H. R. B., & Cerri, C. E. P. (2017) · Carbon Management · doi:10.1080/17583004.2017.1309204 · Cited by 1
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.