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 1,001–1,100 of 4,089
Model‐based variance partitioning for statistical ecology
Schulz, T., Saastamoinen, M., & Vanhatalo, J. P. (2025) · Ecological Monographs · doi:10.1002/ecm.1646 · Cited by 34
Ghosh, P., & Tu, W. (2009) · Journal of the American Statistical Association · doi:10.1198/jasa.2009.0013 · Cited by 28
Bayesian hierarchical modeling of the dynamics of spatio-temporal influenza season outbreaks
Lawson, A. B., & Song, H. (2010) · Spatial and Spatio-temporal Epidemiology · doi:10.1016/j.sste.2010.03.001 · Cited by 29
Ramesh, T., Kalle, R., Rosenlund, H., & Downs, C. T. (2015) · Forest Ecology and Management · doi:10.1016/j.foreco.2015.10.005 · Cited by 30
Black, C. K., Masters, M. D., LeBauer, D. S., Anderson‐Teixeira, K. J., & DeLucia, E. H. (2017) · Plant and Soil · doi:10.1007/s11104-017-3333-7 · Cited by 29
“S1) via Bayesian priors that are Bweakly informative^ in the sense of Gelman (2006).”
MethodGershony, L. C., Belanger, J. M., Short, A. D., Le, M., Hytönen, M. K., Lohi, H., Famula, T. R., Kennedy, L. J., & Oberbauer, A. M. (2019) · Canine Genetics and Epidemiology · doi:10.1186/s40575-019-0070-7 · Cited by 27
Bayesian model selection in linear mixed models for longitudinal data
Ariyo, O., Quintero, A., Muñoz, J., Verbeke, G. N., & Lesaffre, E. M. E. H. (2019) · Journal of Applied Statistics · doi:10.1080/02664763.2019.1657814 · Cited by 38
“The conditionally conjugate prior for the randomeffects covariancematrix is the inverseWishart distribution, but this choice has been shown to be problematic when the number of clusters (here subjects) is small [16,42].”
BackgroundMethodForecasting from others’ experience: Bayesian estimation of the generalized Bass model
Hassan, A. R., & Montoya‐Blandón, S. (2019) · International Journal of Forecasting · doi:10.1016/j.ijforecast.2019.05.016 · Cited by 30
The Choice Architecture of School Choice Websites
Glazerman, S., Nichols-Barrer, I., Valant, J., Chandler, J., & Burnett, A. (2020) · Journal of Research on Educational Effectiveness · doi:10.1080/19345747.2020.1716905 · Cited by 30
“In selecting these priors, we followed previous work (Gelman, 2006) and the current recommendations from the Stan Development Team (2016).”
Sarma, A., & Kay, M. (2020) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3313831.3376377 · Cited by 41
“Normative recommendations state that weakly informative priors should be intentionally weaker than whatever actual prior knowledge is available [15,46], but how much weaker and in what way is not always well-defined.”
BackgroundChoosing increases the value of non-instrumental information
Jiwa, M., Cooper, P. S., Chong, T. T. ‐., & Bode, S. (2021) · Scientific Reports · doi:10.1038/s41598-021-88031-y · Cited by 33
A meta-analysis of event-related potential correlates of recognition memory
Morcom, A., Kwon, S., Rugg, M. D., Curran, T., & Wiegand, R. (2023) · Psychonomic Bulletin & Review · doi:10.3758/s13423-023-02309-y · Cited by 39
“Weakly informative prior distributions were set on the mean effect estimates and the cross-study heterogeneity tau (Gelman, 2006; Harrer et al., 2021; Williams et al., 2018).”
BackgroundDreassi, E., Petrucci, A., & Rocco, E. (2013) · Biometrical Journal · doi:10.1002/bimj.201200271 · Cited by 28
“We use the half-Cauchy distribution, consistently with the recommendations given in Gelman (2006) and Polson and Scott (2012), to achieve non-informativeness for variance parameters.”
MethodA Bayesian approach to probabilistic sensitivity analysis in structured benefit‐risk assessment
Waddingham, E., Mt‐Isa, S., Nixon, R. M., & Ashby, D. (2015) · Biometrical Journal · doi:10.1002/bimj.201300254 · Cited by 33
“…effect, but the overall result remained robust: with a correlation coefficient of 0.9 between all pairs of outcomes (the strongest C© 2015 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.biometrical-journal.com *Half-Cauchy distribution with scale parameter 25; see Gelman (2006); ** where σ = τ−0.5.”
ResultInfluentialObjective priors for hypothesis testing in one‐way random effects models
García‐Donato, G., & Sun, D. (2007) · Canadian Journal of Statistics · doi:10.1002/cjs.5550350207 · Cited by 28
Ghosh, P., Basu, S., & Tiwari, R. C. (2009) · Journal of the American Statistical Association · doi:10.1198/jasa.2009.0038 · Cited by 27
“Gelman (2006) recently questioned the routine use of the inverse gamma prior for the variance parameter and instead recommended a half-t family, this prior can also be considered for the error variance in our joinpoint regression model.”
A generalized Poisson-gamma model for spatially overdispersed data
Neyens, T., Faes, C., & Molenberghs, G. (2011) · Spatial and Spatio-temporal Epidemiology · doi:10.1016/j.sste.2011.10.004 · Cited by 34
An adaptive truncation method for inference in Bayesian nonparametric models
Griffin, J. E. (2014) · Statistics and Computing · doi:10.1007/s11222-014-9519-4 · Cited by 27
“The notation FT (ν, A ) denotes the folded t -distribution which has the density This is a heavy-tailed distribution which was proposed by Gelman (2006) for variance parameters in hierarchical models.”
MethodBayesian spatial modeling of data from avian point count surveys
Webster, R. A., Pollock, K. H., & Simons, T. R. (2008) · Journal of Agricultural Biological and Environmental Statistics · doi:10.1198/108571108x311563 · Cited by 27
“We sampled values from the posterior of τφ = 1/σ 2φ , for which we used a prior density of p(τφ) ∝ τ −3/2 φ , which corresponds to a uniform prior on σφ as recommended by Gelman (2006).”
A statistical model for under- or overdispersed clustered and longitudinal count data
Grunwald, G. K., Bruce, S. L., Jiang, L., et al. (2011) · Biometrical Journal · doi:10.1002/bimj.201000076 · Cited by 27
Applications of Bayesian approach in modelling risk of malaria-related hospital mortality
Kazembe, L. N., Chirwa, T. F., Simbeye, J. S., & Namangale, J. J. (2008) · BMC Medical Research Methodology · doi:10.1186/1471-2288-8-6 · Cited by 30
Estimation of a Nonlinear Intervention Phase Trajectory for Multiple-Baseline Design Data
Hembry, I., Bunuan, R., Beretvas, S. N., Ferron, J. M., & Noortgate, W. V. D. (2014) · The Journal of Experimental Education · doi:10.1080/00220973.2014.907231 · Cited by 29
“Gelman (2006) suggested use of the uniform or half-Cauchy distributions for √ τ for analyzing datasets with a small number of level 2 units.”
BackgroundMethodKrafty, R. T., Rosen, O., Stoffer, D. S., Buysse, D. J., & Hall, M. H. (2017) · Journal of the American Statistical Association · doi:10.1080/01621459.2017.1281811 · Cited by 26
“…are placed by assuming that τβrk , τγ rk , τδrk , τβik , τγ ik , τδik , k > = 1, . . . ,P − 1, τβdkk, τγ dkk, τδdkk, k = 1, . . . ,P, are independent Half-t(ν,G) random variables with pdf p(x) ∝ [1 + (x/G)2/ν)]−(ν+1)/2, x > 0, where the hyperparameters ν and G are assumed known (Gelman 2006).”
The Generalized Multilevel Facets Model for Longitudinal Data
Hung, L.-F., & Wang, W. (2011) · Journal of Educational and Behavioral Statistics · doi:10.3102/1076998611402503 · Cited by 27
“While using a Bayesian approach, the specification of the prior distribution plays a crucial role (Gelman, 2006).”
MethodYou, M., Byon, E., Jin, J. J., et al. (2016) · IISE Transactions · doi:10.1080/0740817x.2016.1204489 · Cited by 27
“Illustrations of the first- and second-order neighborhood turbines of Turbine C. (16), we employ non-informative priors by using the same shape and scale parameters (Gelman, 2006).”
MethodKirch, C., Edwards, M. C., Meier, A., & Meyer, R. (2018) · Bayesian Analysis · doi:10.1214/18-ba1126 · Cited by 28
Bayesian mortality forecasting with overdispersion
Wong, J. S. T., Forster, J. J., & Smith, P. W. F. (2017) · Insurance Mathematics and Economics · doi:10.1016/j.insmatheco.2017.09.023 · Cited by 27
“Since we have no knowledge on the appropriate extent of overdispersion in our data, we assign the conditional conjugate (see Gelman, 2006) prior σ−2µ ∼ Gamma(aµ, bµ), with aµ = bµ = 0.0001 for computational purposes under the PLNLC model.”
MethodInfluentialGabrio, A., Mason, A. J., & Baio, G. (2018) · Statistics in Medicine · doi:10.1002/sim.8045 · Cited by 31
Bonnet, T., Morrissey, M. B., & Kruuk, L. E. B. (2019) · Journal of Heredity · doi:10.1093/jhered/esz018 · Cited by 31
Non‐Gaussian dynamic Bayesian modelling for panel data
Juárez, M. A., & Steel, M. F. J. (2009) · Journal of Applied Econometrics · doi:10.1002/jae.1113 · Cited by 30
“Gelman (2006) argues that the resulting posterior can be quite sensitive to the particular choice of (small) aτ if the data allow mass for τ close to zero.”
BackgroundSparks, P. J., Sparks, C. S., & Campbell, J. J. A. (2012) · GeoJournal · doi:10.1007/s10708-011-9445-3 · Cited by 30
Variable smoothing in Bayesian intrinsic autoregressions
Brewer, M., & Nolan, A. J. (2007) · Environmetrics · doi:10.1002/env.844 · Cited by 31
Seventy-year long record of monthly water balance estimates for Earth’s largest lake system
Do, H. X., Smith, J. P., Fry, L. M., & Gronewold, A. D. (2020) · Scientific Data · doi:10.1038/s41597-020-00613-z · Cited by 30
Meager, R. (2015) · SSRN Electronic Journal · doi:10.2139/ssrn.2620834 · Cited by 40
Performance of evaluation metrics for classification in imbalanced data
Huayanay, A. D. L. C., Bazán, J. L., & Russo, C. M. (2024) · Computational Statistics · doi:10.1007/s00180-024-01539-5 · Cited by 26
“We follow the suggestion of Gelman (2006) and initally consider a non-informative uniform prior density, noting that the use of an inverse-gamma( c , c ) family of non-informative prior distributions is not recommended, since in cases where this parameter is estimated to be approximately 0, the…”
Multivariate Beta Regression with Application in Small Area Estimation
de Souza, D. F., & Moura, F. A. S. (2016) · Journal of Official Statistics · doi:10.1515/jos-2016-0038 · Cited by 26
A bias‐adjusted evidence synthesis of RCT and observational data: the case of total hip replacement
Schnell‐Inderst, P., Iglesias, C. P., Arvandi, M., Ciani, O., Gothe, R. M., Peters, J., Blom, A. W., Taylor, R. S., & Siebert, U. (2017) · Health Economics · doi:10.1002/hec.3474 · Cited by 30
“We used a t-distribution for the mean and three different weakly informative priors for the variance (i.e. Gamma, Inverse-Gamma, and Half-Cauchy distributions), as suggested by Gelman (2006).”
MethodBayesian inference under cluster sampling with probability proportional to size
Makela, S., Si, Y., & Gelman, A. (2018) · Statistics in Medicine · doi:10.1002/sim.7892 · Cited by 35
“3 Prior specification and computation We use the following weakly informative prior distributions as recommended by Gelman (2006), α0, γ0, α1, γ1 ind ∼ N(0, 10) σβ0 , σβ1 , σy ind ∼ Cauchy(0, 2.”
Self-citationJiang, Z., & Skorupski, W. P. (2017) · Behavior Research Methods · doi:10.3758/s13428-017-0986-3 · Cited by 31
“…= c(50,60) is the grand means that users calculated by averaging data points of each subtest; R = diagonal(1) is used to specifying an unstructured uninformative prior distributions for the precision (inverse of covariance matrix), which is assumed to follow a Wishart distribution (Gelman, 2006).”
MethodHow Are Emergent Constraints Quantifying Uncertainty and What Do They Leave Behind?
Williamson, D. B., & Sansom, P. G. (2019) · Bulletin of the American Meteorological Society · doi:10.1175/bams-d-19-0131.1 · Cited by 44
“As argued by (Gelman, 2006), a natural choice of prior for σ is a Half-Normal prior, σ ∼ HN(0, σ2s), (9) where the Half-Normal distribution shares the form of the pdf of the traditional Normal distribution, but with support restricted to σ > 0.”
BackgroundMethodInfluentialLink, R. M., Fuchs, S., Arias‐Aguilar, D., Leuschner, C., Ugalde, M. C., Valverde, J. C., & Schuldt, B. (2020) · Agricultural and Forest Meteorology · doi:10.1016/j.agrformet.2020.107913 · Cited by 26
“All parameters in the model were assigned weakly informative priors (Gelman, 2006; Gelman et al., 2008).”
MethodClimate change-driven elevational changes among boreal nocturnal moths
Keret, N., Mutanen, M., Orell, M. I., Itämies, J., & Välimäki, P. (2020) · Oecologia · doi:10.1007/s00442-020-04632-w · Cited by 28
“Prior sensitivity was analyzed by fitting the same models, but parameterizing the priors so that the observed variance would divide evenly among the variance components with a slight increase in the degree of belief (υ = k + 1, where k is the number of random factors) (Gelman 2006).”
MethodHaemosporidian parasites of Neotropical birds: Causes and consequences of infection
Ellis, V. A., Fecchio, A., & Ricklefs, R. E. (2020) · The Auk · doi:10.1093/auk/ukaa055 · Cited by 26
“We used non-informative priors (specifically, parameter expanded priors [Gelman 2006]) and let the Markov chain run for 1 million iterations for each model; convergence was confirmed by visual inspection of the Markov chains.”
MethodKerioui, M., Mercier, F., Bertrand, J., Tardivon, C., Bruno, R., Guedj, J., & Desmée, S. (2020) · Statistics in Medicine · doi:10.1002/sim.8756 · Cited by 35
Zitzmann, S., Helm, C., & Hecht, M. (2021) · Frontiers in Psychology · doi:10.3389/fpsyg.2020.611267 · Cited by 30
“The choice of prior has received a lot of attention in the methodological literature (e.g., Natarajan and Kass, 2000; Gelman, 2006; Chung et al., 2013), and scholars have made different suggestions about how priors can be specified in advantageous ways.”
BackgroundA meta-analysis of the capitalisation of CAP direct payments into land prices
Varacca, A., Guastella, G., Pareglio, S., & Sckokai, P. (2021) · European Review of Agricultural Economics · doi:10.1093/erae/jbab014 · Cited by 31
Zhang, X., Yan, B., Du, C., Cheng, C., & Zhao, H. (2021) · Atmospheric Pollution Research · doi:10.1016/j.apr.2021.101248 · Cited by 26
Spatiotemporal high-resolution prediction and mapping: methodology and application to dengue disease
Jaya, I. G. N. M., & Folmer, H. (2022) · Journal of Geographical Systems · doi:10.1007/s10109-021-00368-0 · Cited by 28
Korup, O. (2020) · Earth Surface Processes and Landforms · doi:10.1002/esp.4995 · Cited by 26
Thorp, S., & Mandel, K. S. (2022) · Monthly Notices of the Royal Astronomical Society · doi:10.1093/mnras/stac2714 · Cited by 26
Choi, K., & SELTZER, M. H. (2009) · Journal of Educational and Behavioral Statistics · doi:10.3102/1076998609337138 · Cited by 47
Kwon, H., Khalil, A. F., & Siegfried, T. (2008) · JAWRA Journal of the American Water Resources Association · doi:10.1111/j.1752-1688.2008.00173.x · Cited by 28
“This setting gives a much sharper inference, using the partial pooling that comes with fitting a hierarchical model (Gelman, 2006).”
MethodSmith, A., Anderson, M. J., Millar, R., & Willis, T. (2013) · Marine Ecology Progress Series · doi:10.3354/meps10653 · Cited by 29
“The square roots of variance components for random factors were given standard half-Cauchy priors (Gelman 2006).”
MethodSOURCES OF VARIATION IN DETECTION OF WADING BIRDS FROM AERIAL SURVEYS IN THE FLORIDA EVERGLADES
Conroy, M. J., Peterson, J. T., Bass, O. L., Fonnesbeck, C., Howell, J. E., Moore, C. T., & Runge, J. P. (2008) · The Auk · doi:10.1525/auk.2008.07134 · Cited by 26
Demographic analyses of a fox population suffering from sarcoptic mange
Uraguchi, K., Ueno, M., Iijima, H., & Saitoh, T. (2014) · Journal of Wildlife Management · doi:10.1002/jwmg.794 · Cited by 26
Ibáñez, B., Librero, J., Peiró, S., et al. (2011) · BMC Medical Research Methodology · doi:10.1186/1471-2288-11-172 · Cited by 26
Characterizing the information content of a newly hatched chick's first visual object representation
Wood, J. N. (2014) · Developmental Science · doi:10.1111/desc.12198 · Cited by 25
“I used a uniform density kappa (Gelman, 2006) that ranged from 0.000001 (i.e. very little consistency across subjects) to the maximum reasonable kappa.”
MethodInfluentialOn comparing traits and abundance for predicting species interactions with imperfect detection
Weinstein, B. G., & Graham, C. H. (2017) · Food Webs · doi:10.1016/j.fooweb.2017.05.002 · Cited by 26
Selective interspecific information use in the nest choice of solitary bees
Loukola, O., Gatto, E., Híjar-Islas, A. C., & Chıttka, L. (2020) · Animal Biology · doi:10.1163/15707563-20191233 · Cited by 27
“In our model, we used the study site as random effect and a parameter-expanded prior that was minimally informative (Gelman, 2006) (V = 1, ν = 0.”
MethodInfluentialPan, S., Wu, H., Yen, A. M., et al. (2007) · Statistics in Medicine · doi:10.1002/sim.2999 · Cited by 26
Bayesian modeling of multivariate spatial binary data with applications to dental caries
Bandyopadhyay, D., Reich, B. J., & Slate, E. H. (2009) · Statistics in Medicine · doi:10.1002/sim.3647 · Cited by 25
Flexible distributions for triple-goal estimates in two-stage hierarchical models
Paddock, S. M., Ridgeway, G. K., Lin, R., & Louis, T. A. (2005) · Computational Statistics & Data Analysis · doi:10.1016/j.csda.2005.05.008 · Cited by 26
How Many Data Points and How Large an R-Squared Value is Essential for Arrhenius Plots?
Taira, K., McInnes, D., & Zhang, L. (2022) · Journal of Catalysis · doi:10.1016/j.jcat.2023.01.033 · Cited by 28
Carroll, J. M., Yeager, D. S., Buontempo, J., Hecht, C. A., Cimpian, A., Mhatre, P. C., Muller, C. L., & Crosnoe, R. L. (2023) · Monographs of the Society for Research in Child Development · doi:10.1111/mono.12471 · Cited by 42
Calder, C. A., Craigmile, P. F., Cressie, N., Li, H., & Paul, R. (2009) · Bayesian Analysis · doi:10.1214/09-ba401 · Cited by 25
Ball, S. J., Haynes, A., Jacoby, P. A., Pereira, G. F., Miller, L., Bower, C. I., & Davis, E. A. (2014) · Health & Place · doi:10.1016/j.healthplace.2014.05.004 · Cited by 27
Sources of Uncertainty in Future Projections of the Carbon Cycle
Hewitt, A. J., Booth, B. B. B., JONES, C. D., Robertson, E., Wiltshire, A., Sansom, P. G., Stephenson, D. B., & Yip, S. (2016) · Journal of Climate · doi:10.1175/jcli-d-16-0161.1 · Cited by 27
“The half-Cauchy distribution with scale parameter A has the following form: The half-Cauchy distribution has the advantage of concentrating less of the prior mass close to zero than the inverse gamma distribution (Gelman 2006).”
BackgroundInfluentialIsotocin neuronal phenotypes differ among social systems in cichlid fishes
Reddon, A. R., O’Connor, C. M., Nesjan, E., Cameron, J., Hellmann, J. K., Ligocki, I. Y., Marsh‐Rollo, S. E., Hamilton, I. M., WYLIE, D. R. W., Hurd, P. L., et al. (2017) · Royal Society Open Science · doi:10.1098/rsos.170350 · Cited by 31
Detecting Multiple Random Changepoints in Bayesian Piecewise Growth Mixture Models
Lock, E. F., Kohli, N., & Bose, M. (2018) · Psychometrika · doi:10.1007/s11336-017-9594-5 · Cited by 28
“001)) is generally not recommended for modeling the variance of hierarchical random effects (Gelman et al., 2006), partly because its density is unstable as the variance approaches 0.”
BackgroundMethodBrandt, H., Cambria, J., & Kelava, A. (2018) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705511.2018.1474114 · Cited by 26
“Typical distributions used are the inverse gamma distribution or the Half Cauchy distribution (Gelman, 2006; Polson & Scott, 2011).”
MethodWang, Z., Yue, Y., He, B., Nie, K., Tu, W., Du, Q., & Li, Q. (2020) · International Journal of Geographical Information Systems · doi:10.1080/13658816.2020.1798967 · Cited by 25
Bayesian analysis for uncertainty quantification of in situ stress data
Feng, Y., Bozorgzadeh, N., & Harrison, J. P. (2020) · International Journal of Rock Mechanics and Mining Sciences · doi:10.1016/j.ijrmms.2020.104381 · Cited by 25
Jiang, L., Nie, L., Yan, F., & Yuan, Y. (2021) · Contemporary Clinical Trials · doi:10.1016/j.cct.2021.106460 · Cited by 29
“It is useful to reduce and solve the common problem of blindly using a large-variance inverse-gamma prior or a prior arbitrarily specified, which may lead to pathological statistical properties (Gelman, 2006).”
MethodSilva, L. D., Elias, R. B., & Silva, L. (2021) · Environmental Modelling & Software · doi:10.1016/j.envsoft.2021.105203 · Cited by 28
Anthropogenic land‐use change shapes bird diversity along the eastern Himalayan altitudinal gradient
Penjor, U., Jamtsho, R., & Sherub, S. (2021) · Journal of Applied Ecology · doi:10.1111/1365-2664.14101 · Cited by 28
Ren, W., Yu, B., Chen, Y., & Gao, K. (2022) · International Journal of Environmental Research and Public Health · doi:10.3390/ijerph191811358 · Cited by 35
“A well-known computing approach for Bayesian inference, the Markov chain Monte Carlo (MCMC) method [55], was used in this study to better approximate the target posterior distribution.”
MethodHegedus, P. B., Maxwell, B. D., & Mieno, T. (2022) · Precision Agriculture · doi:10.1007/s11119-022-09968-2 · Cited by 27
“Thus, Bayesian MLR models vary in their covariates between fields and splits but are all fit from the same initial model; where predictor variables are covariates from Table 3 with uniform priors on coefficients 0 − 16 (Gelman, 2006), and the SAR spatial random error, z , defined in Eq.”
BackgroundEstimating the prevalence of social and emotional loneliness across the adult lifespan
Manoli, A., McCarthy, J., & Ramsey, R. (2022) · Scientific Reports · doi:10.1038/s41598-022-24084-x · Cited by 35
Intuitive joint priors for Bayesian linear multilevel models: The R2D2M2 prior
Aguilar, J. E., & Bürkner, P. (2023) · Electronic Journal of Statistics · doi:10.1214/23-ejs2136 · Cited by 36
“We follow the recommendations of Gelman (2006) to consider a half Student-t prior on σ with ν degrees of freedom and scale η.”
MethodTwo of a Kind: Comparing Big and Small Black Holes in Binaries with Gravitational Waves
Farah, A. M., Fishbach, M., & HOLZ, D. E. (2024) · The Astrophysical Journal · doi:10.3847/1538-4357/ad0558 · Cited by 50
Zhou, S., Punt, A. E., Deng, R. A., et al. (2009) · Marine and Freshwater Research · doi:10.1071/mf09022 · Cited by 25
Sacchi, E., & Sayed, T. (2014) · Accident Analysis & Prevention · doi:10.1016/j.aap.2014.06.016 · Cited by 27
Profiting from pilot studies: Analysing mortality using Bayesian models with informative priors
Morris, W. K., Vesk, P. A., & Mccarthy, M. A. (2012) · Basic and Applied Ecology · doi:10.1016/j.baae.2012.11.003 · Cited by 28
Claassen, C. (2015) · Scientometrics · doi:10.1007/s11192-015-1584-8 · Cited by 30
“Given that the number of rating systems included is not large (P ¼ 8), I use more 5 Glockner-Rist and Hoijtink (2009) describes the parallels between IRT and FA models. informative half-Cauchy ð0; 2:5Þ prior distributions for the intercept and slope variance parameters, r2 (Gelman 2006).”
MethodA Bayesian nonparametric meta‐analysis model
Karabatsos, G., Talbott, E., & Walker, S. G. (2014) · Research Synthesis Methods · doi:10.1002/jrsm.1117 · Cited by 28
“Furthermore, we specify uniform prior density un(σ0|0, b0) for the varianceσ20 of the random intercepts μ0j ∞ j¼ ∞, for a reasonably large value b0, as consistent with previous recommendations (Gelman, 2006, Section 7.1).”
MethodResultInfluentialDiez, J. M., Buckley, H. L., Case, B. S., et al. (2009) · Journal of Applied Ecology · doi:10.1111/j.1365-2664.2009.01725.x · Cited by 25
A Poisson Regression Model to Examine Spatial Patterns in Antenatal Care Utilisation in Nigeria
Gayawan, E. (2013) · Population Space and Place · doi:10.1002/psp.1775 · Cited by 25
Kuusela, M., & Panaretos, V. M. (2015) · The Annals of Applied Statistics · doi:10.1214/15-aoas857 · Cited by 25
“Unfortunately, the resulting estimates are known to be sensitive to this nontrivial choice [Gelman (2006)].”
BackgroundMcEntee, J. P. (2014) · Behavioral Ecology · doi:10.1093/beheco/aru136 · Cited by 27
“A half-Cauchy distribution was used as a prior in the random-effects variances to restrict sampling space to reasonable values (Gelman 2006), and parameter expansion (van Dyk and Meng 2001; Gelman 2006) and slice sampling were employed (Damien et al.”
MethodInvestigating the use of formant frequencies in listener judgments of speaker size
Barreda, S. (2015) · Journal of Phonetics · doi:10.1016/j.wocn.2015.11.004 · Cited by 28
Smith, S. J., Sameoto, J. A., & Brown, C. J. (2016) · Canadian Journal of Fisheries and Aquatic Sciences · doi:10.1139/cjfas-2015-0595 · Cited by 28
Approximation of Bayesian Predictive p-Values with Regression ABC
Nott, D. J., Drovandi, C., Mengersen, K. L., et al. (2016) · Bayesian Analysis · doi:10.1214/16-ba1033 · Cited by 25
How proper are Bayesian models in the astronomical literature?
Tak, H., Ghosh, S. K., & Ellis, J. A. (2018) · Monthly Notices of the Royal Astronomical Society · doi:10.1093/mnras/sty2326 · Cited by 27
“For a second-level variance component in a Gaussian hierarchical model such as σ2 in Equation (2), Gelman (2006) does not recommend an inverse-Gamma(a, b) prior with arbitrarily small values for both a and b as a non-informative choice7.”
BackgroundBlonder, B. W., Salinas, N., Bentley, L. P., et al. (2018) · Journal of Ecology · doi:10.1111/1365-2745.12945 · Cited by 25
“Prior distributions for each effect were chosen as inverse gamma with location and scale parameters set to 0.01 (Gelman, 2006).”
MethodInfluentialHan, X., Li, J., & Wang, N. (2018) · BMC Public Health · doi:10.1186/s12889-018-5417-6 · Cited by 25
“According to the conclusion of Gelman (2006) [20], the prior distribution of mean square error for all random variables in the model is determined as a strictly positive half-Gaussian distributionN+∞(0, 10).”
MethodBayesian functional joint models for multivariate longitudinal and time-to-event data
Li, K., & Luo, S. (2018) · Computational Statistics & Data Analysis · doi:10.1016/j.csda.2018.07.015 · Cited by 33
Three certified sugar reference materials for carbon isotope delta measurements
Chartrand, M. M. G., Meija, J., Kumkrong, P., & Mester, Z. (2018) · Rapid Communications in Mass Spectrometry · doi:10.1002/rcm.8357 · Cited by 25
Single-trial EEG dissociates motivation and conflict processes during decision-making under risk
Pornpattananangkul, N., Grogans, S. E., Yu, R., & Nusslock, R. (2018) · NeuroImage · doi:10.1016/j.neuroimage.2018.12.029 · Cited by 29
“We employed standard normal and half-Cauchy prior distributions for group-level means (μ) and standard deviations (σ), respectively (Gelman, 2006).”
Rosenbloom, D. I. S., Bacchetti, P., Stone, M., Deng, X., Bosch, R. J., Richman, D. D., Siliciano, J. D., Mellors, J. W., Deeks, S. G., Ptak, R. G., et al. (2019) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1006849 · Cited by 26
“A uniform or normal distribution was used as prior for each fixed-effect parameter, and a mixture of a point-mass at zero and a half-Cauchy prior was used for each random effect parameter; this prior has a heavy tail, allowing the model to explore large values of σ [16, 17], while also allowing it to take on the null…”
MethodAdaptation reduces competitive dominance and alters community assembly
Nadeau, C. P., Farkas, T. E., Makkay, A. M., Papke, R. T., & Urban, M. C. (2021) · Proceedings of the Royal Society B Biological Sciences · doi:10.1098/rspb.2020.3133 · Cited by 24
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.