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 901–1,000 of 4,089
Famula, R. A., Richards, J. H., Famula, T. R., & Neale, D. B. (2018) · Tree Genetics & Genomes · doi:10.1007/s11295-018-1307-4 · Cited by 41
A Bayesian Multivariate Functional Dynamic Linear Model
Kowal, D. R., & Matteson, D. S. (2014) · Cited by 31
Evanno, G., Castella, E., & Goudet, J. (2006) · Journal of Evolutionary Biology · doi:10.1111/j.1420-9101.2006.01098.x · Cited by 36
“We also tested Uniform distributions (0, 10 000) as priors as it was recently suggested that these distributions may have better frequentist properties than Gamma distributions (Gelman, 2006; O’Hara & Merilä, 2005).”
MethodProbabilistic Meshless Methods for Partial Differential Equations and Bayesian Inverse Problems.Preprint
Cockayne, J., Oates, C. J., Sullivan, T. J., et al. (2016) · arXiv (Cornell University) · Cited by 31
Heath, A., Manolopoulou, I., & Baio, G. (2019) · Medical Decision Making · doi:10.1177/0272989x19837983 · Cited by 35
Historical reconstruction of the Puget Sound (USA) groundfish community
Essington, T., Ward, E., Francis, T., Greene, C. H., Kuehne, L. M., & Lowry, D. (2020) · Marine Ecology Progress Series · doi:10.3354/meps13547 · Cited by 31
López‐Baucells, A., Rowley, S., Rocha, R., Bobrowiec, P. E. D., Palmeirim, J. M., Farneda, F. Z., & Meyer, C. F. J. (2022) · Landscape Ecology · doi:10.1007/s10980-022-01493-x · Cited by 33
“All models were fitted with a Gaussian error distribution and a “non-informative” prior, which is weakly informative and is equivalent to an inverse-gamma prior with shape and scale equal to 0.001 (Gelman 2006; Wilson et al.”
MethodInfluentialUnderstanding memory impairment with memory models and hierarchical Bayesian analysis
Pooley, J. P., Lee, M. D., & Shankle, W. R. (2010) · Journal of Mathematical Psychology · doi:10.1016/j.jmp.2010.08.003 · Cited by 28
“Following the general advice of Gelman (2006), the standard deviations of the discriminability and response bias parameter distributions for stage i are both given non-informative uniform priors, with σd′,i, σc,i ∼ Uniform(5/100, 3).”
MethodBayesian latent variable models for the analysis of experimental psychology data
Merkle, E. C., & Wang, T. (2016) · Psychonomic Bulletin & Review · doi:10.3758/s13423-016-1016-7 · Cited by 35
“As a result, we use “weakly informative” prior distributions on the parameters (Gelman, 2006): prior distributions with variances constrained to reflect plausible parameter values.”
MethodBayesian Piecewise Linear Mixed Models With a Random Change Point
Brilleman, S. L., Howe, L. D., Wolfe, R. S., et al. (2017) · Epidemiology · doi:10.1097/ede.0000000000000723 · Cited by 30
Lian, Q., Hodges, J. S., & Chu, H. (2018) · Journal of the American Statistical Association · doi:10.1080/01621459.2018.1476239 · Cited by 38
“However, this strategy can be computationally challenging for large K. Alternative parameterizations could be helpful, for example, Daniels and Kass (1999), Gelman (2006), or the covariance matrices could be simplified, as demonstrated in the case study.”
BackgroundMallard resource selection trade‐offs in a heterogeneous environment during autumn and winter
Palumbo, M. D., Petrie, S. A., Schummer, M. L., Rubin, B. D., & Bonner, S. J. (2019) · Ecology and Evolution · doi:10.1002/ece3.4864 · Cited by 31
Huh, D., Mun, E.-Y., Walters, S., Zhou, Z.-Y., & Atkins, D. C. (2019) · Addictive Behaviours · doi:10.1016/j.addbeh.2019.01.032 · Cited by 15
“For the SD of the varying intercepts, we used a half-Cauchy distribution with a mode of 0 and a shape parameter of 25, as suggested by Gelman (2006) for variance terms in multilevel models.”
MethodShades of dark uncertainty and consensus value for the Newtonian constant of gravitation
Merkatas, C., Toman, B., Possolo, A. M., & Schlamminger, S. (2019) · Metrologia · doi:10.1088/1681-7575/ab3365 · Cited by 40
“, n; (ii) G has an essentially non-informative Gaussian prior distribution with mean G2014 and standard deviation 100 times larger than u(G2014); (iii) the {τk} have mildly informative half-Cauchy distributions [23] whose medians are either the median absolute deviation from the median of the {u(Gj )} (when K = 1), or…”
BackgroundBarbasch, T. A., Rueger, T., Srinivasan, M., Wong, M. Y. L., Jones, G. P., & Buston, P. M. (2020) · Oikos · doi:10.1111/oik.07674 · Cited by 35
Mitsova, D., Esnard, A., Sapat, A. K., et al. (2020) · Natural Hazards Review · doi:10.1061/(asce)nh.1527-6996.0000421 · Cited by 30
Urbina, D. A., & Rodrı́guez, G. (2021) · Journal of Economic Studies · doi:10.1108/jes-05-2020-0199 · Cited by 30
Association between productivity and journal impact across disciplines and career age
Sunahara, A. S., Perc, M., & Ribeiro, H. V. (2021) · Physical Review Research · doi:10.1103/physrevresearch.3.033158 · Cited by 36
Two-vehicle driver-injury severity: A multivariate random parameters logit approach
Gong, H., Fu, T., Sun, Y., Guo, Z., Lin, C., Hu, W., & Ling, Z. (2021) · Analytic Methods in Accident Research · doi:10.1016/j.amar.2021.100190 · Cited by 35
Hierarchical Bayesian model for predicting small-strain stiffness of sand
Tao, Y., Phoon, K., Sun, H., & Cai, Y. (2023) · Canadian Geotechnical Journal · doi:10.1139/cgj-2022-0598 · Cited by 31
Lucka, F. F. (2012) · Inverse Problems · doi:10.1088/0266-5611/28/12/125012 · Cited by 30
“An advantageous choice for p pr ( σ 2 ) is given by the conjugate prior w.r.t. p li ( m | u, σ 2 ) , which is the inverse gamma distribution [19]: with shape and scale parameters α and β ( Γ denotes the gamma function).”
Extinction under a behavioral microscope: Isolating the sources of decline in operant response rate
Cheung, T. H. C., Neisewander, J. L., & Sanabria, F. (2012) · Behavioural Processes · doi:10.1016/j.beproc.2012.02.012 · Cited by 34
Feedback and Modularization in a Bayesian Meta–analysis of Tree Traits Affecting Forest Dynamics
Ogle, K., Barber, J., & Sartor, K. (2013) · Bayesian Analysis · doi:10.1214/13-ba806 · Cited by 30
Román, J. C., & Hobert, J. P. (2012) · The Annals of Statistics · doi:10.1214/12-aos1052 · Cited by 25
“) The standard diffuse prior for this model, which is among those recommended by Gelman (2006), has density 1/ ( σ2 e √ σ2 α ) .”
BackgroundNg, E. S. W., Diaz‐Ordaz, K., Grieve, R., Nixon, R. M., Thompson, S. G., & Carpenter, J. R. (2016) · Statistical Methods in Medical Research · doi:10.1177/0962280213511719 · Cited by 32
Individual heterogeneity and senescence in Silvereyes on Heron Island
Knape, J., Jonzén, N., Sköld, M., et al. (2010) · Ecology · doi:10.1890/10-0183.1 · Cited by 30
Bayesian semiparametric model with spatially–temporally varying coefficients selection
Cai, B., Lawson, A. B., Hossain, M. M., Choi, J., Kirby, R. S., & Liu, J. (2013) · Statistics in Medicine · doi:10.1002/sim.5789 · Cited by 30
Vaughan, A. S., Kramer, M. R., Waller, L. A., et al. (2015) · Annals of Epidemiology · doi:10.1016/j.annepidem.2015.02.007 · Cited by 30
Nguyen, C. M., Bounthavong, M., Mendes, M., Christopher, M. L. D., Tran, J. N., Kazerooni, R., & Morreale, A. P. (2012) · PharmacoEconomics · doi:10.2165/11594990-000000000-00000 · Cited by 32
Almeling, L., Sennhenn‐Reulen, H., Hammerschmidt, K., Freund, A. M., & Fischer, J. (2017) · American Journal of Primatology · doi:10.1002/ajp.22711 · Cited by 34
“Typically, the data provides enough information to estimate β0, β1, . . . by the use of any reasonable non-informative prior distribution (Gelman, 2006).”
BackgroundMethodBayesian Energy Measurement and Verification Analysis
Carstens, H., Xia, X., & Yadavalli, V. S. S. (2018) · Energies · doi:10.3390/en11020380 · Cited by 37
Cross-Shelf Variation in Coral Community Response to Disturbance on the Great Barrier Reef
Mellin, C., Thompson, A., Jonker, M. J., & Emslie, M. J. (2019) · Diversity · doi:10.3390/d11030038 · Cited by 30
An Overview of the U.S. Census Bureau's Small Area Income and Poverty Estimates Program
Bell, W. R., Basel, W. W., & Maples, J. J. (2016) · doi:10.1002/9781118814963.ch19 · Cited by 32
Disease implications of animal social network structure: a synthesis across social systemsPreprint
Sah, P., Mann, J., & Bansal, S. (2017) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/106633 · Cited by 31
“We used a weakly informative Gelman prior for fixed effects and parameterexpanded priors for the random effects to improve mixing and decrease the autocorrelation among iterations (Gelman, 2006).”
MethodEffect of spatial connectivity on host resistance in a highly fragmented natural pathosystem
Höckerstedt, L., Sirén, J., & Laine, A. (2018) · Journal of Evolutionary Biology · doi:10.1111/jeb.13268 · Cited by 31
Risk factors for prolonged air leak after pulmonary surgery: A systematic review and meta-analysis
Zheng, Q., Ge, L., Zhou, J. H., Zhang, Y., Lyu, M., Chen, C., Wang, T., & Liu, L. (2022) · Asian Journal of Surgery · doi:10.1016/j.asjsur.2022.01.001 · Cited by 39
Tourism Demand in the Face of Geopolitical Risk: Insights From a Cross-Country Analysis
Papagianni, E., Evgenidis, A., Tsagkanos, A. G., & Megalooikonomou, V. (2023) · Journal of Travel Research · doi:10.1177/00472875231206539 · Cited by 51
Cornulier, T., Robinson, R. A., Elston, D. A., Lambin, X., Sutherland, W. J., & Benton, T. G. (2010) · Methods in Ecology and Evolution · doi:10.1111/j.2041-210x.2010.00054.x · Cited by 37
Robust linear mixed models for Small Area Estimation
Fabrizi, E., & Trivisano, C. (2009) · Journal of Statistical Planning and Inference · doi:10.1016/j.jspi.2009.07.022 · Cited by 33
Meta-analysis of functional neuroimaging data using Bayesian nonparametric binary regression
Yu, Y., Lindquist, M. A., & Loh, J. M. (2012) · The Annals of Applied Statistics · doi:10.1214/11-aoas523 · Cited by 33
“This expanded model form allows conditionally conjugate prior distributions for both ξ and θ, and these parameters are independent in the conditional posterior distribution [Gelman (2006); Gelman et al.”
MethodInfluentialKamarainen, A. M., Rowland, F. E., Biggs, R., et al. (2008) · Canadian Journal of Fisheries and Aquatic Sciences · doi:10.1139/f08-161 · Cited by 29
Images from a jointly-arousing collective ritual reveal affective polarization
Bulbulia, J. A., Xygalatas, D., Schjoedt, U., Fondevila, S., Sibley, C. G., & Konvalinka, I. (2013) · Frontiers in Psychology · doi:10.3389/fpsyg.2013.00960 · Cited by 32
“We used parameter expanded priors (Hadfield, 2010, p. 128) as recommended by Gelman (2006), however use of inverse-Wishart priors for the random effects variances made no practical difference to theoretical inferences, as the posterior distribution of parameter co/variances were very similar.”
MethodMcParland, D., Baron, S., O’Rourke, S., Dowling, D. P., Ahearne, E., & Parnell, A. C. (2017) · Journal of Intelligent Manufacturing · doi:10.1007/s10845-017-1317-3 · Cited by 31
“We use a Bayesian hierarchical model (Gelman and Hill 2006) to simultaneously learn the toolwear rates in each force direction whilst constraining them in an R2 random field according to their cutting speeds and feed rate.”
MethodThe relational luring effect: Retrieval of relational information during associative recognition.
Popov, V., Hristova, P., & Anders, R. (2017) · Journal of Experimental Psychology General · doi:10.1037/xge0000305 · Cited by 40
“The priors for the fixed effects were uninformative normal distributions centered around 0 with SD=10(8), and the random effects had uninformative Cauchy priors with a scale of 25 (Gelman, 2006).”
BackgroundEffects of rotational grazing management on nesting greater sage‐grouse
Smith, J. T., Tack, J. D., Berkeley, L. I., et al. (2017) · Journal of Wildlife Management · doi:10.1002/jwmg.21344 · Cited by 29
Grüebler, M. U., Müller, M. S., Michel, V. T., Perrig, M., Keil, H., Naef‐Daenzer, B., & Korner‐Nievergelt, F. (2018) · Animal Behaviour · doi:10.1016/j.anbehav.2018.05.009 · Cited by 34
Normington, J., Zhu, J., Mattiello, F., et al. (2019) · Contemporary Clinical Trials · doi:10.1016/j.cct.2019.105890 · Cited by 29
Shinohara, Y., Osaka, T., Inoue, I., Iwashita, T., Dmowski, W., Ryu, C. W., Sarathchandran, Y., & Egami, T. (2020) · Nature Communications · doi:10.1038/s41467-020-20036-z · Cited by 35
Ahmed, I., Gaweesh, S. M., & Ahmed, M. M. (2020) · Transportation Research Record Journal of the Transportation Research Board · doi:10.1177/0361198120931103 · Cited by 30
“5) priors for the regression coefficients and HalfCauchy (0, 1) for the random intercept standard deviation (31, 46, 47).”
BackgroundModeling Item Revisit Behavior: The Hierarchical Speed–Accuracy–Revisits Model
Bezirhan, U., von Davier, M., & Grabovsky, I. (2020) · Educational and Psychological Measurement · doi:10.1177/0013164420950556 · Cited by 34
“For the inverse of the item time discrimination parameter weakly informative half-Cauchy distribution, 1=ai;Cauchy+(0, 4), is preferred because it is more flexible and performs better near the origin compared with the inverse gamma family (Gelman, 2006).”
BackgroundTatt‐BiLSTM: Web service classification with topical attention‐based BiLSTM
Kang, G., Xiao, Y., Liu, J., Cao, Y., Cao, B., Zhang, X., & Ding, L. (2021) · Concurrency and Computation Practice and Experience · doi:10.1002/cpe.6287 · Cited by 27
“HDP, 13 an extension of LDA, is a multilayer form of the Dirichlet process, designed to address cases in topic document modeling where the number of topic terms is not known in advance.”
BackgroundUnbalanced Risk of Pulmonary Tuberculosis in China at the Subnational Scale: Spatiotemporal Analysis
Hu, M., Feng, Y., Li, T., Zhao, Y., Wang, J., Xu, C., & Chen, W. (2022) · JMIR Public Health and Surveillance · doi:10.2196/36242 · Cited by 37
“As suggested by Gelman [20], a strictly positive half Gaussian prior N+∞(0, 10) was assigned to parameters of random-effect standard deviations, including σv and σε.”
MethodInfluentialNot what u expect: Effects of prediction errors on item memory.
Ortiz-Tudela, J., Nolden, S., Pupillo, F., Ehrlich, I., Schommartz, I., Turan, G., & Shing, Y. L. (2023) · Journal of Experimental Psychology General · doi:10.1037/xge0001367 · Cited by 38
Hierarchical modeling in spatial epidemiology
Lawson, A. B. (2014) · Wiley Interdisciplinary Reviews Computational Statistics · doi:10.1002/wics.1315 · Cited by 29
Feng, C. X., & Dean, C. B. (2012) · Environmetrics · doi:10.1002/env.2158 · Cited by 28
“…large value of variances; the priors assigned to the variances of the spatially structured and unstructured random effects 2 b ; 2 h1 ; ; 2 hJ can be inverse Gamma distributions or their standard deviations can be uniformly distributed, because of the robust properties of this prior (Gelman, 2006).”
BackgroundLucas, R. W., Sponseller, R. A., & Laudon, H. (2013) · Ecosystems · doi:10.1007/s10021-013-9641-8 · Cited by 33
“Finally, we assigned a uniform prior between 0 and 10,000 to the observation standard deviation (Gelman 2006).”
MethodModelling individual fertility levels in Malawian women: a spatial semiparametric regression model
Kazembe, L. N. (2007) · Statistical Methods & Applications · doi:10.1007/s10260-007-0076-2 · Cited by 32
Dwyer, J. M., Fensham, R. J., & Buckley, Y. M. (2010) · Ecological Applications · doi:10.1890/09-1690.1 · Cited by 29
Cubaynes, S., Doutrelant, C., Grégoire, A., Perret, P., Faivre, B., & Giménez, O. (2011) · Ecology · doi:10.1890/11-0258.1 · Cited by 33
Similarity in G matrix structure among natural populations ofArabidopsis lyrata
Puentes, A., Granath, G., & Ågren, J. (2016) · Evolution · doi:10.1111/evo.13034 · Cited by 28
“To examine the robustness of the posterior distribution to the prior (Gelman 2006; Teplitsky et al. 2014a), we ran the models with different priors (V = diag(n) × 0.1, V = diag(n) × 10), and nu = n) and obtained very similar results.”
ResultInfluentialDiversification across a heterogeneous landscape
Walter, G. M., Wilkinson, M. J., James, M. E., Richards, T. J., Aguirre, J. D., & Ortíz-Barrientos, D. (2016) · Evolution · doi:10.1111/evo.13009 · Cited by 32
Electoral Competitiveness and Turnout in British Elections, 1964–2010
Vowles, J., Katz, G., & Stevens, D. P. (2015) · Political Science Research and Methods · doi:10.1017/psrm.2015.67 · Cited by 30
AGE-DEPENDENT SURVIVAL ANALYZED WITH BAYESIAN MODELS OF MARK–RECAPTURE DATA
Zheng, C., Ovaskainen, O., Saastamoinen, M., et al. (2007) · Ecology · doi:10.1890/06-1246.1 · Cited by 29
Efficient and Scalable Bayesian Neural Nets with Rank-1 FactorsPreprint
Dusenberry, M. W., Jerfel, G., Wen, Y., Ma, Y.-A., Snoek, J., Heller, K., Lakshminarayanan, B., & Tran, D. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2005.07186 · Cited by 260
“Weakly informative priors such as the Cauchy are often preferred for robustness as they concentrate less probability at the mean thanks to heavier tails (Gelman et al., 2006).”
BackgroundAbbruzzini, T. F., Moreira, M. Z., de Camargo, P. B., Conz, R. F., & Cerri, C. E. P. (2017) · Agricultural Research · doi:10.1007/s40003-017-0281-7 · Cited by 29
“For the proportion between parameters CL and CR, it was assigned a scale non-informative beta distribution prior, and for the variance parameters, it was assigned to half-cauchy on the standard deviation scale and the respective random effects simulated by their conditional expectation [12, 13].”
MethodMak, M., Faber, M., & Willems, R. M. (2023) · Cortex · doi:10.1016/j.cortex.2023.01.014 · Cited by 34
“As variance can only be positive, weakly regularizing, half-cauchy priors with a mean of 0 and a standard deviation of 1 were used for the variance of the random effects as well as the overall variance (as suggested by Gelman, 2006; McElreath, 2016).”
Camac, J., Williams, R. J., Wahren, C.-H., Jarrad, F., Hoffmann, A. A., & Vesk, P. A. (2015) · Oecologia · doi:10.1007/s00442-015-3261-2 · Cited by 29
Modern Methods of Estimating Biodiversity from Presence-Absence Surveys
Robert, M., Nicholas, J., & Aaron, M. (2011) · InTech eBooks · doi:10.5772/23881 · Cited by 64
“Each of the heterogeneity parameters ( σ a 0 , σ b 0 , σ b l ) was assigned a half-Cauchy prior (Gelman 2006) with unit scale parameter, which has probability density function Gelman (2006) showed that this prior avoids problems that can occur when alternative “noninforma-tive” priors are used…”
InfluentialImproved latent space approach for modelling non-stationary spatial–temporal random fields
Xu, H., & Gardoni, P. (2018) · Spatial Statistics · doi:10.1016/j.spasta.2018.01.003 · Cited by 28
Hennig, E. I., & Mazzi, D. (2018) · Insects · doi:10.3390/insects9030118 · Cited by 28
BART with targeted smoothing: An analysis of patient-specific stillbirth risk
Starling, J. E., Murray, J. S., Carvalho, C. M., Bukowski, R. K., & Scott, J. G. (2020) · The Annals of Applied Statistics · doi:10.1214/19-aoas1268 · Cited by 35
Predicting a Distal Outcome Variable From a Latent Growth Model: ML versus Bayesian Estimation
Smid, S. C., Depaoli, S. A., & van de Schoot, R. (2019) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705511.2019.1604140 · Cited by 33
“Options to further explore the behavior of the variance parameters under varying prior distributions include the use of the half-Cauchy prior as suggested by Gelman (2006), or the use of reference priors as speci fi ed in Tsai and Hsiao (2008).”
Bucur, I. G., Claassen, T., & Heskes, T. (2019) · Statistical Methods in Medical Research · doi:10.1177/0962280219851817 · Cited by 34
“X, Y : For the scale parameters, we choose a weakly informative, but proper half-Gaussian prior with a large standard deviation, as suggested by Gelman.64”
MethodBayesian Model Selection in Fisheries Management and Ecology
Doll, J. C., & Jacquemin, S. (2019) · Journal of Fish and Wildlife Management · doi:10.3996/042019-jfwm-024 · Cited by 28
Rushing, C. S., Rohrbaugh, R. W., Fiss, C. J., Rosenberry, C. S., Rodewald, A. D., & Larkin, J. L. (2019) · Forest Ecology and Management · doi:10.1016/j.foreco.2019.117629 · Cited by 34
Pleban, J. R., Guadagno, C. R., Mackay, D. S., Weinig, C., & Ewers, B. E. (2020) · PLANT PHYSIOLOGY · doi:10.1104/pp.19.00375 · Cited by 31
“Priors for the precision terms used weakly informed normal distributions (Gelman, 2006).”
BackgroundKim, T., Kwon, H., & Kim, K. B. (2020) · Journal of Hydrology · doi:10.1016/j.jhydrol.2020.125790 · Cited by 30
Schmidt, J., & Moss, R. E. S. (2021) · Computers and Geotechnics · doi:10.1016/j.compgeo.2020.103963 · Cited by 30
Kameyama, R., Nishihara, G. N., Kawagoe, C., & Terada, R. (2021) · Journal of Applied Phycology · doi:10.1007/s10811-021-02469-1 · Cited by 28
“Informative normal priors were placed on all parameters of the model by using values from a previous study (Kokubu et al. 2015), and a halfCauchy prior distribution was placed on the scale parameter of the models (Gelman 2004, 2006).”
MethodPapaix, J. J., Cubaynes, S., Buoro, M., Charmantier, A., Perret, P., & Giménez, O. (2010) · Journal of Evolutionary Biology · doi:10.1111/j.1420-9101.2010.02079.x · Cited by 33
“We assigned uniform distributions to the standard deviation of the random effects, rt U[0,10], re U[0,10] and ra U[0,10] (Gelman, 2006; Royle, 2008).”
MethodCircular conditional autoregressive modeling of vector fields
Modlin, D., Fuentes, M., & Reich, B. J. (2011) · Environmetrics · doi:10.1002/env.1133 · Cited by 28
“Gelman (2006) discussed the possible prior distributions that could be utilized for scale parameters.”
Torman, V. B. L., & Camey, S. A. (2015) · Emerging Themes in Epidemiology · doi:10.1186/s12982-015-0030-y · Cited by 33
Treatment Monitoring of HIV‐Infected Patients based on Mechanistic Models
Prague, M., Commenges, D., Drylewicz, J., & Thiébaut, R. (2012) · Biometrics · doi:10.1111/j.1541-0420.2012.01749.x · Cited by 30
“Half-Cauchy priors were taken for the variance of the random effects, as recommended by Gelman (2006).”
MethodMa, J., Thabane, L., Kaczorowski, J., Chambers, L., Dolovich, L., Karwalajtys, T., & Levitt, C. (2009) · BMC Medical Research Methodology · doi:10.1186/1471-2288-9-37 · Cited by 28
Ecologically realistic estimates of maximum population growth using informed B ayesian priors
Delean, S., Brook, B. W., & Bradshaw, C. J. A. (2012) · Methods in Ecology and Evolution · doi:10.1111/j.2041-210x.2012.00252.x · Cited by 30
Finding the features that represent stimuli
Zeigenfuse, M. D., & Lee, M. D. (2009) · Acta Psychologica · doi:10.1016/j.actpsy.2009.07.014 · Cited by 28
“To specify prior distributions, we followed the advice of Gelman (2006) and placed a uniform prior on the variance”
MethodDerbridge, J. J., Merkle, J. A., Bucci, M. E., Callahan, P., Koprowski, J. L., Polfus, J. L., & Krausman, P. R. (2015) · PLoS ONE · doi:10.1371/journal.pone.0119940 · Cited by 36
“If posterior distributions more closely resemble non-informative priors, the selected informative prior distribution may be problematic [47, 48].”
BackgroundPark, J., Abdel-Aty, M. A., Wang, J.-H., et al. (2015) · Accident Analysis & Prevention · doi:10.1016/j.aap.2015.03.025 · Cited by 28
Prague, M., Commenges, D., Guedj, J., Drylewicz, J., & Thiébaut, R. (2013) · Computer Methods and Programs in Biomedicine · doi:10.1016/j.cmpb.2013.04.014 · Cited by 36
“If we assume normal independent priors for the fixed effects, half-Cauchy priors for the variances of the random effects, as recommended by Gelman (2006) with median parameter s j 2 and conventional one-dimension Jeffreys-type improper priors for the variances of the measurement errors, the…”
MethodPrey and Habitat Influences the Movement of Clapper Rails in Northern Gulf Coast Estuaries
Rush, S. A., Mordecai, R. S., Woodrey, M. S., & Cooper, R. J. (2010) · Waterbirds · doi:10.1675/063.033.0317 · Cited by 32
Bayesian estimation of the network autocorrelation model
Dittrich, D., Leenders, R., & Mulder, J. (2016) · Social Networks · doi:10.1016/j.socnet.2016.09.002 · Cited by 28
“We fitted the model in a Bayesian framework in R using standard non-informative uniform priors for µ ρ , τ ρ , and log σ 2 ρ (Gelman, 2006).”
Kuehn, N. M., & Abrahamson, N. (2017) · Bulletin of the Seismological Society of America · doi:10.1785/0120170166 · Cited by 31
Understanding How People use Search to Support their Everyday Creative Tasks
Zhang, Y., & Capra, R. (2019) · Conference on Human Information Interaction and Retrieval · doi:10.1145/3295750.3298936 · Cited by 36
Intuitive Joint Priors for Variance Parameters
Fuglstad, G., Hem, I. G., Knight, A., Rue, H., & Riebler, A. (2019) · Bayesian Analysis · doi:10.1214/19-ba1185 · Cited by 40
Ancelet, S., Abellan, J. J., Vilas, V. D. R., Birch, C., & Richardson, S. T. (2012) · Biometrical Journal · doi:10.1002/bimj.201000106 · Cited by 38
“…, σ 2 v1, and σ 2 v2 and IGamma(0.01,0.01) for σ 2 2, respectively; (ii) one set of lefttruncated normal priors with mean 0 and variance 100 for both σu, σv1, σv2, and σ 2; and (iii) one set of uniform priors on [0, 10] for σu, σv1, σv2, and [0, 100] for σ 2 inspired in the work by Gelman (2006).”
BackgroundInfluentialHeritability of Asymmetry and Lateral Plate Number in the Threespine Stickleback
Loehr, J., Leinonen, T., Herczeg, G., O’Hara, R. B., & Merilä, J. (2012) · PLoS ONE · doi:10.1371/journal.pone.0039843 · Cited by 27
Conditional predictive inference for online surveillance of spatial disease incidence
Vallet, A. C., & Lawson, A. B. (2011) · Statistics in Medicine · doi:10.1002/sim.4340 · Cited by 30
The Performance of Random Coefficient Regression in Accounting for Residual Confounding
Gustafson, P., & Greenland, S. (2006) · Biometrics · doi:10.1111/j.1541-0420.2005.00510.x · Cited by 29
“In some instances other families of priors for variance components may be preferred (see, for instance, Gelman, 2006), but the inverse-gamma specification is convenient for illustrative purposes.”
BackgroundAdamczyk, A., Kim, C., & Schmuhl, M. A. (2017) · Sociological Perspectives · doi:10.1177/0731121417724563 · Cited by 30
Tremlett, C. J., Cleasby, I. R., Bolton, M., & Wilson, L. J. (2024) · Bird Study · doi:10.1080/00063657.2024.2438641 · Cited by 35
“Uniform priors for standard deviations were chosen instead of those constructed using inverse-gamma distributions because the former are considered less informative (Gelman 2006).”
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.