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 1,201–1,300 of 4,089
Sorting your way out: Perceived party positions, political knowledge, and polarization
Vegetti, F., Fazekas, Z., & Méder, Z. Z. (2016) · Acta Politica · doi:10.1057/s41269-016-0029-6 · Cited by 22
Bringing It All Together: Multi-species Integrated Population Modelling of a Breeding Community
Lahoz‐Monfort, J. J., Harris, M. P., Wanless, S., et al. (2017) · Journal of Agricultural Biological and Environmental Statistics · doi:10.1007/s13253-017-0279-4 · Cited by 22
“…priors N ( 0, 104 ) for resight probabilities p∗s (in the logit scale); (iv) flat proper priors U (0, 3) for SDs of random terms (σδφ, σδρ, σεφs, σερs), following Gelman (2006); and (v) flat proper priors for SD σxs of the observation errors: U (0, 15000) for puffins, U (0, 5000) for the…”
Madlon‐Kay, S., Montague, M. J., Brent, L. J. N., Ellis, S., Zhong, B. L., Snyder‐Mackler, N., Horvath, J. E., Skene, J. H. P., & Platt, M. L. (2018) · American Journal of Primatology · doi:10.1002/ajp.22873 · Cited by 21
“Finally, the fixed effects, random effects terms are given weaklyinformative priors (Gelman, 2006):”
Skelsey, P., Humphris, S. N., Campbell, E., et al. (2018) · PLoS ONE · doi:10.1371/journal.pone.0205711 · Cited by 22
Environment predicts repeated body size shifts in a recent radiation of Australian mammals*
Roycroft, E., Nations, J. A., & Rowe, K. C. (2019) · Evolution · doi:10.1111/evo.13859 · Cited by 23
“We fit regularizing normal priors on the population-level effects (N (0, 1.5)) to prevent MCMC chains from occasionally searching very large, unreasonable values of model space (Gelman 2006; McElreath 2016).”
MethodInfluentialKasper, C., Ruiz‐Ascacibar, I., Stoll, P., et al. (2020) · Journal of Animal Breeding and Genetics · doi:10.1111/jbg.12472 · Cited by 22
“We ran the models with weakly informative parameter-expanded priors (Gelman, 2006) with a variance of 1 for each efficiency trait and each variance component as well as the residual, and a degree-of-belief parameter equal to the dimensions of the variance–covariance matrix (V = diag(2), nu = 2, alpha.”
MethodDuncan, E. W., & Mengersen, K. L. (2020) · PLoS ONE · doi:10.1371/journal.pone.0233019 · Cited by 29
“Other hyperpriors are possible (see Gelman [36] for example), but are not considered here for the sake of brevity.”
BackgroundBayesian Forecasting of Mortality Rates by Using Latent Gaussian Models
Alexopoulos, A., Δελλαπόρτας, Π., & Forster, J. J. (2018) · UCL Discovery (University College London) · doi:10.17863/cam.34547 · Cited by 22
Polson, N. G., & Scott, J. G. (2010) · Cited by 20
Green, T. R. F., Murphy, S. M., Ortiz, J. B., & Rowe, R. K. (2022) · Frontiers in Neurology · doi:10.3389/fneur.2021.804139 · Cited by 28
“Specifically, we applied ∼Normal (0, 1) priors to population-level parameters, ∼Cauchy (0, 5) priors to the variance scale parameters, and ∼Cauchy (0, 2) priors to the standard deviations of group-level effects, thereby appropriately restricting those variance components’ parameter spaces to positive values (42).”
MethodCapanu, M., Orlow, I., Berwick, M., Hummer, A., Thomas, D. C., & Begg, C. B. (2008) · Statistics in Medicine · doi:10.1002/sim.3196 · Cited by 22
Defining informative priors for ensemble modeling in systems biology
Tsigkinopoulou, A., Hawari, A., Uttley, M., & Breitling, R. (2018) · Nature Protocols · doi:10.1038/s41596-018-0056-z · Cited by 21
Feller, A., Grindal, T., Miratrix, L. W., & Page, L. C. (2014) · SSRN Electronic Journal · doi:10.2139/ssrn.2534811 · Cited by 21
European Institutional Integration and the Educational Divide in Support for the European Union
Baute, S., & Tober, T. (2024) · British Journal of Political Science · doi:10.1017/s0007123424000309 · Cited by 21
Spatial patterning and correlates of self-harm in Manchester, England
Lin, C., Bickley, H., Clements, C., Webb, R. T., Gunnell, D., Hsu, C., Chang, S., & Kapur, N. (2019) · Epidemiology and Psychiatric Sciences · doi:10.1017/s2045796019000696 · Cited by 25
“Non-informative prior distributions were used in the estimation of the Bayesian models; for example, the standard errors of unstructured and structured variability were specified using a uniform distribution (0, 5) in the analysis (Gelman, 2006).”
MethodInfluentialTime‐varying relationships between ocean conditions and sockeye salmon productivity
Malick, M. J. (2020) · Fisheries Oceanography · doi:10.1111/fog.12469 · Cited by 24
Bunji, K., & Okada, K. (2020) · Behavior Research Methods · doi:10.3758/s13428-019-01302-5 · Cited by 23
“For the priors of the item parameters, we adopted the half−t distribution for ξ j and β jk as weakly informative priors, which are recommended by Gelman (2006) and Stan Development Team (2017).”
MethodHaq, M. T., Zlatkovic, M., & Ksaibati, K. (2021) · Accident Analysis & Prevention · doi:10.1016/j.aap.2021.106392 · Cited by 22
Explaining declines of newt abundance in northern Italy
Falaschi, M., Muraro, M., Gibertini, C., Monache, D. D., Parrino, E. L., Faraci, F., Belluardo, F., Nicola, M. R. D., Manenti, R., & Ficetola, G. F. (2022) · Freshwater Biology · doi:10.1111/fwb.13909 · Cited by 25
Ogero, M. O., Orwa, J., Odhiambo, R., Agoi, F., Lusambili, A. M., Obure, J., Temmerman, M., Luchters, S. M. F., & Ngugi, A. K. (2022) · BMC Public Health · doi:10.1186/s12889-022-12570-w · Cited by 27
Parks, R. M., Nunez, Y., Balalian, A. A., Gibson, E. A., Hansen, J., Raaschou‐Nielsen, O., Ketzel, M., Khan, J., Brandt, J. G., Vermeulen, R. C. H., et al. (2022) · Epidemiology · doi:10.1097/ede.0000000000001536 · Cited by 25
Keller, B. T., & Enders, C. K. (2023) · Multivariate Behavioral Research · doi:10.1080/00273171.2022.2147049 · Cited by 24
Covariate handling approaches in combination with dynamic borrowing for hybrid control studies
Fu, C., Pang, H., Zhou, S., & Zhu, J. (2023) · Pharmaceutical Statistics · doi:10.1002/pst.2297 · Cited by 31
“argues that the half-Cauchy prior is more flexible and has better performance when the number of groups is small.(37) This prior distribution allows most of the variance parameters to be small but with the occasional large variance.”
BackgroundEarthquake forecasting from paleoseismic records
Wang, T., Griffin, J., Brenna, M., Fletcher, D., Zeng, J., Stirling, M., Dillingham, P. W., & Kang, J. (2024) · Nature Communications · doi:10.1038/s41467-024-46258-z · Cited by 27
Murray, T. A., Hobbs, B. P., Sargent, D. J., & Carlin, B. P. (2015) · Bayesian Analysis · doi:10.1214/15-ba954 · Cited by 26
Liess, A., Guo, J., Lind, M. I., et al. (2015) · Journal of Animal Ecology · doi:10.1111/1365-2656.12426 · Cited by 21
Monitoring and statistical modelling of sedimentation in gully pots
Post, J. A. B., Pothof, I., Dirksen, J., et al. (2015) · Water Research · doi:10.1016/j.watres.2015.10.021 · Cited by 21
Bayesian Estimation of Log-Normal Means with Finite Quadratic Expected Loss
Fabrizi, E., & Trivisano, C. (2012) · Bayesian Analysis · doi:10.1214/12-ba733 · Cited by 21
Meta‐Analysis of Studies with Missing Data
Yuan, Y., & Little, R. J. A. (2008) · Biometrics · doi:10.1111/j.1541-0420.2008.01068.x · Cited by 23
“In the case that k is very low, the half-Cauchy prior may be a useful alternative (Gelman, 2006).”
BackgroundBayesian hypothesis testing for pattern discrimination in brain decoding
Olivetti, E., Veeramachaneni, S., & Nowakowska, E. (2011) · Pattern Recognition · doi:10.1016/j.patcog.2011.04.025 · Cited by 22
Lecomte, J.-B., Benoît, H. P., Étienne, M., Bel, L., & Parent, É. (2013) · Ecological Modelling · doi:10.1016/j.ecolmodel.2013.06.017 · Cited by 27
McParland, D., Phillips, C. M., Brennan, L., Roche, H. M., & Gormley, I. C. (2017) · Statistics in Medicine · doi:10.1002/sim.7371 · Cited by 29
“Sensitivity to these inverse gamma hyperparameter values is a known problem for Bayesian inference of models of this type Gelman (2006).”
Evans, J. M., Noorai, R. E., Tsai, K. L., Starr-Moss, A. N., Hill, C. M., Anderson, K. J., Famula, T. R., & Clark, L. A. (2017) · PLoS Genetics · doi:10.1371/journal.pgen.1006604 · Cited by 26
“Subsequently, we consider the weakly informative prior for σ ~ Cauchy(0,25) [70].”
BackgroundIijima, H. (2017) · Wildlife Biology · doi:10.2981/wlb.00329 · Cited by 21
“Prior distributions of s3 and s4 were a uniform distribution from 0 to 100 (Gelman 2006).”
BackgroundSpatial small area smoothing models for handling survey data with nonresponse
Watjou, K., Faes, C., Lawson, A. T. D.-D., Kirby, R. S., Aregay, M., Carroll, R. M., & Vandendijck, Y. (2017) · Statistics in Medicine · doi:10.1002/sim.7369 · Cited by 22
Masri, S., Li, L., Dang, A., Chung, J. M., Chen, J., Fan, Z. T., & Wu, J. Y. (2018) · Atmospheric Environment · doi:10.1016/j.atmosenv.2018.01.014 · Cited by 22
“This formula illustrates that any sample point within a certain region r will be affected by their neighboring regions (r′ ∈ σr). τ2 is a priori variance that is weakly informatively determined using inverse gamma distribution (Gelman, 2006).”
BackgroundZero‐inflated spatio‐temporal models for disease mapping
Torabi, M. (2017) · Biometrical Journal · doi:10.1002/bimj.201600120 · Cited by 21
“We also use uniform distribution U (0, 1000) instead of the gamma distribution as a prior for the standard deviations (Gelman, 2006).”
MethodNitulescu, R., Young, J., Saeed, S., Cooper, C., Cox, J., Martel‐Laferrière, V., Hull, M., Walmsley, S. L., TYNDALL, M. W., Wong, A., et al. (2018) · International Journal of Drug Policy · doi:10.1016/j.drugpo.2018.08.012 · Cited by 22
Han, T., Zhu, G., Ma, J., Wang, S., Zhang, K., Liu, X., Ma, T., Shang, S., & Huang, C. (2019) · Photosynthesis Research · doi:10.1007/s11120-019-00684-z · Cited by 24
“For the variance parameters, we chose folded Cauchy densities to generate the priors (Gelman 2004, 2006).”
MethodYamanaka, S., Yamaura, Y., Sayama, K., Sato, S., & Ozaki, K. (2021) · Forest Ecology and Management · doi:10.1016/j.foreco.2021.119073 · Cited by 22
Decision Tree for Key Comparisons
Possolo, A. M., Koepke, A., Newton, D., & Winchester, M. R. (2021) · Journal of Research of the National Institute of Standards and Technology · doi:10.6028/jres.126.007 · Cited by 34
Tan, Y. V., Flannagan, C. A. C., & Elliott, M. R. (2018) · Statistics and Its Interface · doi:10.4310/sii.2018.v11.n4.a1 · Cited by 21
“, ˙2 and (T j;M j)s are independent. We assume that 2 ˘IG(1 2 ;1 2 ) and take B !1when evaluating the posterior distribution of ˘. This eectively turns the FNCT prior on ˝2 to a half-Cauchy prior (Gelman, 2006). The posterior draws of a k and ˝ can be obtained by setting a k = ˘ k and ˝= j˘j. The at improper prior and…”
BackgroundNathoo, F. S., & Ghosh, P. (2012) · Statistics in Medicine · doi:10.1002/sim.5504 · Cited by 21
Combining data from multiple sources, with applications to environmental risk assessment
RYAN, L. M. (2007) · Statistics in Medicine · doi:10.1002/sim.3053 · Cited by 22
Zhou, S., Punt, A. E., Deng, R. A., et al. (2011) · Canadian Journal of Fisheries and Aquatic Sciences · doi:10.1139/f2011-052 · Cited by 20
Herrero, A., Almaraz, P., Zamora, R., Castro, J., & Hódar, J. A. (2015) · Journal of Ecology · doi:10.1111/1365-2745.12527 · Cited by 20
“Uniform distributions were placed on the standard deviations of the environmental and demographic noise terms in the Gompertz state-space model, and in the terms for environmental and process parameter noise in eqns 3–5: rnt , d, rht , rxi;t , qbk;t ;t , qck;t ;t, qrk;t ;t Uð0; 3Þ; see Gelman 2006).”
MethodBuscot, M., Wotherspoon, S. J., Magnussen, C. G., Juonala, M., Sabin, M. A., Burgner, D. P., Lehtimäki, T. J., Viikari, J. S. A., Hutri‐Kähönen, N., Raitakari, O. T., et al. (2017) · BMC Medical Research Methodology · doi:10.1186/s12874-017-0358-9 · Cited by 22
Robustly estimating the marginal likelihood for cognitive models via importance sampling
Tran, M., Scharth, M., Gunawan, D., Kohn, R., Brown, S. D., & Hawkins, G. E. (2020) · Behavior Research Methods · doi:10.3758/s13428-020-01348-w · Cited by 24
“The standard deviation parameters have half-Cauchy priors (Gelman, 2006).”
BackgroundQuick, M. (2019) · Journal of Geographical Systems · doi:10.1007/s10109-019-00305-2 · Cited by 23
“The standard deviation of each set of random effects terms ( u, 1 , 2 , 3 , , and ) was assigned a positive half-Gaussian prior distribution, Normal+∞ (0, 10) (Gelman 2006).”
BackgroundMethodTrouvé, R., Bunyavejchewin, S., & Baker, P. J. (2020) · Journal of Ecology · doi:10.1111/1365-2745.13343 · Cited by 25
“We used weakly informative priors to help convergence (Gelman, 2006).”
MethodBaek, E., & Ferron, J. M. (2020) · Methodology · doi:10.5964/meth.2817 · Cited by 21
“Similarly, Gelman (2006) found that when the uniform distribution (noninformative prior) is constructed for the standard deviation (SD) unit rather than variance unit of level-2 error, the uniform distribution generally performs well, as long as J ≥ 3 which is required to ensure a proper posterior…”
BackgroundInfluentialZeng, S., Li, F., & Ding, P. (2020) · Journal of the Royal Statistical Society Series A (Statistics in Society) · doi:10.1111/rssa.12595 · Cited by 27
Muir, J. B., & Tkalčić, . (2020) · Geophysical Journal International · doi:10.1093/gji/ggaa397 · Cited by 17
“…the statistical community suggests non-informative priors, such as the Jeffries prior, are in fact poor choices for hyperparameters as they tend to bias hyperparameters to extreme values, and that more restrictive priors such as the half-Cauchy or even half-Normal are better choices (Gelman 2006).”
BackgroundA Bayesian Approach for Uncertainty Quantification in Overcoring Stress Estimation
Feng, Y., Harrison, J. P., & Bozorgzadeh, N. (2020) · Rock Mechanics and Rock Engineering · doi:10.1007/s00603-020-02295-w · Cited by 22
“2015; Gelman 2006; Lemoine 2019; Gelman et al.”
Amer, F. M. A., Thompson, C. G., & Lin, L. (2021) · International Journal of Environmental Research and Public Health · doi:10.3390/ijerph18073492 · Cited by 24
“The uniform prior, denoted by U(0, c), is also widely used for the heterogeneity standard deviation τ [51], where c determines the prior’s upper bound.”
BackgroundDalpasso, A., Ficetola, G. F., Giachello, S., Parrino, E. L., Manenti, R., Muraro, M., & Falaschi, M. (2022) · Diversity and Distributions · doi:10.1111/ddi.13483 · Cited by 21
“For the two random effects, we used a half- Cauchy prior (Gelman, 2006), while for all other parameters, we used a vague normal prior with mean = 0 and variance = 100.”
MethodBertolini, R., Finch, S. J., & Nehm, R. H. (2023) · Frontiers in Education · doi:10.3389/feduc.2023.1073829 · Cited by 32
“The gamma distribution is a continuous probability distribution that is traditionally used as a prior distribution for the variance when nested priors are used (Gelman, 2006).”
Heubner, L., Oertel, R., Tiebel, O., Mehlig-Warnecke, N., Beyer‐Westendorf, J., Mirus, M., Roessler, M., Renner, B., & Spieth, P. M. (2023) · Anesthesiology · doi:10.1097/aln.0000000000004787 · Cited by 21
Nishino, T. (2023) · International Journal of Disaster Risk Reduction · doi:10.1016/j.ijdrr.2023.104124 · Cited by 20
The relationship between possum density and browse damage on kamahi in New Zealand forests
Duncan, R. P., Holland, E. P., Pech, R. P., et al. (2011) · Austral Ecology · doi:10.1111/j.1442-9993.2010.02229.x · Cited by 20
Shelton, A. O., Dick, E. J., Pearson, D. E., et al. (2012) · Canadian Journal of Fisheries and Aquatic Sciences · doi:10.1139/f2011-152 · Cited by 20
“…several prior distributions, we used gamma-distributed priors for all variance parameters, thus providing weakly informative prior information (Gelman 2006; see also Supplementary materials).1 Hierarchical models are easier to sample by treating the qij as unobserved, latent variables and…”
BackgroundMethodBello, N. M., Steibel, J. P., & Tempelman, R. J. (2010) · Biometrical Journal · doi:10.1002/bimj.200900182 · Cited by 20
“Bounded uniform priors are also commonly considered (Sorensen and Gianola, 2002) as, typically, enough data is available to infer upon elements of b i with any reasonable non-informative prior distribution in large field studies (Gelman, 2006).”
BackgroundMethodModels for Bounded Systems with Continuous Dynamics
Cangelosi, A. R., & Hooten, M. B. (2009) · Biometrics · doi:10.1111/j.1541-0420.2008.01130.x · Cited by 20
“Variance components can be modeled in different ways (Gelman, 2006), though the specification used here yielded a stable algorithm and performed well.”
BackgroundJoint prior distributions for variance parameters in Bayesian analysis of normal hierarchical models
Demirhan, H., & Kalaylıoğlu, Z. (2015) · Journal of Multivariate Analysis · doi:10.1016/j.jmva.2014.12.013 · Cited by 20
Generalised linear mixed model analysis via sequential Monte Carlo sampling
Fan, Y., Leslie, D. S., & WAND, M. P. (2008) · Electronic Journal of Statistics · doi:10.1214/07-ejs158 · Cited by 26
A meta-analytic assessment of a Thyroglobulin marker for marbling in beef cattle
Wood, I., Moser, G., Burrell, D. L., et al. (2006) · Genetics Selection Evolution · doi:10.1186/1297-9686-38-5-479 · Cited by 20
Scheuerell, M. D., Buhle, E. R., Semmens, B. X., Ford, M. J., Cooney, T., & Carmichael, R. W. (2015) · Ecology and Evolution · doi:10.1002/ece3.1509 · Cited by 23
“Following Gelman (2006), we assigned noninformative Unif( 100, 100) and Unif(0, 100) priors to the mean (mb) and standard deviation (c), respectively, of the random effects.”
MethodInfluentialGuglielmi, A., Ieva, F., Paganoni, A. M., et al. (2013) · Journal of the Royal Statistical Society Series C (Applied Statistics) · doi:10.1111/rssc.12021 · Cited by 19
“We refer to Gelman (2006) for a discussion on priors of the variance components in hierarchical models.”
MethodVerde, P. E., Ohmann, C., Morbach, S., & Icks, A. (2015) · Statistics in Medicine · doi:10.1002/sim.6809 · Cited by 25
Using Bayesian analysis in repeated preclinical in vivo studies for a more effective use of animals
Walley, R. J., Sherington, J., Rastrick, J., Detrait, E., Hanon, É., & Watt, G. (2016) · Pharmaceutical Statistics · doi:10.1002/pst.1748 · Cited by 23
Posterior Predictive Checks for Conditional Independence Between Response Time and Accuracy
Bolsinova, M. A., & Tijmstra, J. (2016) · Journal of Educational and Behavioral Statistics · doi:10.3102/1076998616631746 · Cited by 20
Bayesian approach to epsilon-skew-normal family
Maleki, M., & Nematollahi, A. R. (2016) · Communication in Statistics- Theory and Methods · doi:10.1080/03610926.2016.1157186 · Cited by 20
Doogan, N., & Warren, K. L. (2017) · Addiction Research & Theory · doi:10.1080/16066359.2016.1249864 · Cited by 21
“A uniform prior is used for the random effect variances as suggested by Gelman (2006).”
MethodStewart, D. R., Butler, M. J., Harris, G. M., et al. (2017) · Biological Conservation · doi:10.1016/j.biocon.2017.02.010 · Cited by 20
The Impact of Prior Information on Bayesian Latent Basis Growth Model Estimation
Shi, D., & Tong, X. (2017) · SAGE Open · doi:10.1177/2158244017727039 · Cited by 24
“However, using noninformative priors should only be seen as a “provisional” strategy and the posterior distribution should always be examined after the model is fitted (Gelman, 2006).”
BackgroundInfluentialIfelebuegu, A. O., Awotu-Ukiri, E. O., Theophilus, S. C., et al. (2017) · Process Safety and Environmental Protection · doi:10.1016/j.psep.2017.10.019 · Cited by 20
“The data obtained from the various databases is directly used for PFD beta distribution [27].”
MethodCombining historical data and bookmakers’ odds in modelling football scores
Egidi, L., Pauli, F., & Torelli, N. (2018) · Statistical Modelling · doi:10.1177/1471082x18798414 · Cited by 30
“7) (1)On the choice of the half-Cauchy distribution for scale parameters, see Gelman et al. (2006).”
BackgroundInfluentialOn the estimation of stress strength reliability parameter of inverted gamma distribution
Iranmanesh, A., Vajargah, K. F., & Hasanzadeh, M. (2018) · Mathematical sciences · doi:10.1007/s40096-018-0246-4 · Cited by 20
Iyengar, S., Lee, S., Irwin, D., et al. (2018) · ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD) · doi:10.1145/3219819.3219825 · Cited by 20
Ouédraogo, M., Kangoye, D. T., Samadoulougou, S., Rouamba, T., Donnen, P., & Kirakoya‐Samadoulougou, F. (2020) · International Journal of Environmental Research and Public Health · doi:10.3390/ijerph17061840 · Cited by 21
“, we assume that all the standard deviations of random effects 1 τφ and σε follow a strictly positive prior semi-Gaussian distribution N+∞(0, 10) [50].”
BackgroundShi, L., Chu, H., & Lin, L. (2020) · Research Synthesis Methods · doi:10.1002/jrsm.1415 · Cited by 23
“37 The uniform prior may be used for τ; alternatively, the inverse-gamma pror or the half-normal prior may be considered.(38,39) Derived from a large database of meta-analyses, evidence-based informative log-normal priors are also available for τ(2).”
MethodA Bayesian framework for describing and predicting the stochastic demand of home care patients
Argiento, R., Guglielmi, A., Lanzarone, E., et al. (2014) · Flexible Services and Manufacturing Journal · doi:10.1007/s10696-014-9200-4 · Cited by 20
“As for as the standard deviation parameters are concerned, a standard noninformative choice (Gelman 2006) is to assume a uniform prior over a large interval ð0; rmaxÞ and then to perform a robustness analysis, decreasing rmax as long as the same posterior inference is obtained.”
Pooling systematic reviews of systematic reviews: a Bayesian panoramic meta‐analysis
Hemming, K., Bowater, R. J., & Lilford, R. J. (2011) · Statistics in Medicine · doi:10.1002/sim.4372 · Cited by 22
A random effects modelling approach to the crossing-fibre problem in tractography
King, M. D., Gadian, D. G., & Clark, C. A. (2008) · NeuroImage · doi:10.1016/j.neuroimage.2008.09.058 · Cited by 21
“A specification based on a uniformly distributed standard deviation has been recommended in preference to the traditional inverse gamma prior distribution with small parameter values (see, for example, Gelman, 2006).”
Goungounga, J. A., Gaudart, J., Colonna, M., & Giorgi, R. (2016) · BMC Medical Research Methodology · doi:10.1186/s12874-016-0228-x · Cited by 21
“The third level (or priors) represents the variability of the spatial process components and is approached with a particular distribution (Beta, Gamma, Dirichlet, or Wishart) [30].”
BackgroundPan, I., & Bester, D. W. (2017) · IEEE Transactions on Fuzzy Systems · doi:10.1109/tfuzz.2017.2746064 · Cited by 22
Shirley, R. B., Smidts, C. S., & Zhao, Y. (2019) · Reliability Engineering & System Safety · doi:10.1016/j.ress.2019.03.023 · Cited by 20
Comment on article by Browne and Draper
Lambert, P. C. (2006) · Bayesian Analysis · doi:10.1214/06-ba117c · Cited by 20
Boehm, U., Annis, J., Frank, M. J., et al. (2018) · doi:10.31234/osf.io/km28u · Cited by 20
Luong, N.-D. M., Coroller, L., Zagorec, M., Moriceau, N., Anthoine, V., Guillou, S., & Membré, J. (2022) · Foods · doi:10.3390/foods11081114 · Cited by 25
“Firstly, Bayesian methods can help to fill data gaps with prior information, possibly provided by experts or other preliminary works [18,19].”
BackgroundKaplan, D. E., Chen, J., Yavuz, S., & Lyu, W. (2022) · Psychometrika · doi:10.1007/s11336-022-09869-3 · Cited by 35
A Bayesian Approach for Data-Driven Dynamic Equation Discovery
North, J. S., Wikle, C. K., & Schliep, E. M. (2022) · Journal of Agricultural Biological and Environmental Statistics · doi:10.1007/s13253-022-00514-1 · Cited by 19
Berman, J., Francoz, D., Abdallah, A., Dufour, S., & Buczinski, S. (2022) · Journal of Dairy Science · doi:10.3168/jds.2021-21695 · Cited by 24
“1) for precision specification was considered as reasonably non-informative (Gelman, 2006).”
BackgroundMethodInfluentialLiu, S., Dupas, R., Guo, D., Lintern, A. M., Minaudo, C., Bende‐Michl, U., Zhang, K., & Duvert, C. (2022) · Water Resources Research · doi:10.1029/2022wr032365 · Cited by 21
“where σβ followed an uninformative uniform distribution between 0 and 10 (σβ ∼ U (0, 10)) (Gelman, 2006; Gelman et al., 2013).”
BackgroundMethodBayesian Analysis of Lifetime Delayed Degradation Process for Destructive/Nondestructive Inspection
Chen, S., Lu, Z., Hu, Q., et al. (2023) · IEEE Transactions on Reliability · doi:10.1109/tr.2023.3330288 · Cited by 19
“In contrast to noninformative priors, weakly informative priors place the probability quality on the reasonable parameter values, while reducing the weight of the nonsensical values [39], [40].”
BackgroundAuthier, M., Martin, C., Ponchon, A., Steelandt, S., Bentaleb, I., & Guinet, C. (2011) · Methods in Ecology and Evolution · doi:10.1111/j.2041-210x.2011.00162.x · Cited by 19
“For the parameter a3, we used a Student-t prior [with location, scale and degrees of freedom set to 0, 10 and 7 respectively (Gelman et al. 2008)]. For modelling V, we used the priors similar to those of Chen & Dunson (2003): independent half-normal priors of mean0 and standard deviation1Æ5 for the elements, cp 2 [1:8…”
MethodBayesian meta‐analysis of Cronbach's coefficient alpha to evaluate informative hypotheses
Okada, K. (2015) · Research Synthesis Methods · doi:10.1002/jrsm.1155 · Cited by 23
“…the prior on the variance parameter τ2, although Brannick and Zhang (2013) use inverse gamma prior on τ2, several authors, including Daniels (1999) and Gelman (2006), recommend against using inverse gamma prior because the inference would consequently be sensitive to the choice of hyperparameters.”
MethodPaddock, S. M. (2014) · Health Services Research · doi:10.1111/1475-6773.12149 · Cited by 19
“A is selected to be large enough to cover all plausible values of the standard deviation of the bk’s (Gelman 2006).”
BackgroundBrun, M., Abraham, C., Jarry, M., Dumas, J., Lange, F., & Prévost, É. (2011) · Ecological Modelling · doi:10.1016/j.ecolmodel.2010.10.030 · Cited by 26
Berejikian, B., Hard, J. J., Tatara, C. P., Doornik, D. V., Swanson, P., & Larsen, D. A. (2017) · Canadian Journal of Fisheries and Aquatic Sciences · doi:10.1139/cjfas-2016-0175 · Cited by 16
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.