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 1,501–1,600 of 4,089
Garre, A., Pielaat, A., Zwietering, M. H., den Besten, H. M. W., & Smid, J. (2022) · International Journal of Food Microbiology · doi:10.1016/j.ijfoodmicro.2022.109935 · Cited by 15
Unethical choice in negotiations: A meta-analysis on gender differences and their moderators
Nohe, C., Hüffmeier, J., Bürkner, P., Mazei, J., Sondern, D., Runte, A., Sieber, F., & Hertel, G. (2022) · Organizational Behavior and Human Decision Processes · doi:10.1016/j.obhdp.2022.104189 · Cited by 17
Predictive ratio CUSUM (PRC): A Bayesian approach in online change point detection of short runs
Bourazas, K., Sobas, F., & Tsiamyrtzis, P. (2023) · Journal of Quality Technology · doi:10.1080/00224065.2022.2161434 · Cited by 15
Qian, J., Yue, M., Huang, P., et al. (2023) · Frontiers in Cellular and Infection Microbiology · doi:10.3389/fcimb.2023.1115087 · Cited by 15
“The last term ϵit eN(0,s 2ϵ ) (Gelman, 2006) is the Gaussian random noise variable and captures additional variability not yet explained by other model components.”
BackgroundAbe, H., Garre, A., Koseki, S., et al. (2023) · Food Control · doi:10.1016/j.foodcont.2023.109831 · Cited by 15
Brown, H. A., Topham, T. H., Clark, B., Ioannou, L. G., Flouris, A. D., Smallcombe, J. W., Telford, R. D., Jay, O., & Périard, J. D. (2023) · Sports Medicine · doi:10.1007/s40279-023-01972-4 · Cited by 20
“In addition, normal distribution priors were set on the standard deviations of group-level effects; core temperature and sweat rate models had a standard deviation of 0.5, whilst the two heart rate models had a standard deviation set at 10 [61].”
InfluentialDrivers of morphological evolution in the toothed whale jaw
Coombs, E. J., Knapp, A., Park, T., Bennion, R. F., McCurry, M. R., Lanzetti, A., Boessenecker, R. W., & McGowen, M. R. (2023) · Current Biology · doi:10.1016/j.cub.2023.11.056 · Cited by 13
“We assessed convergence of the chains by checking the trace plots and then the effective sample size (ESS) with Gelman and Rubin’s diagnostics 98,99 using, respectively, the functions ‘effectiveSize’ and ‘Gelman.diag’ implemented in the R coda package v. 0.19-4 100 (Figure S4).”
MethodInfluentialOn the hierarchical Bayesian modelling of frequency response functions
DARDENO, T. A., Worden, K., Dervilis, N., Mills, R. S., & Bull, L. A. (2024) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2023.111072 · Cited by 26
Bayesian Methods for Information Borrowing in Basket Trials: An Overview
Zhou, T., & Ji, Y. (2024) · Cancers · doi:10.3390/cancers16020251 · Cited by 16
“Three prior choices are considered for the γ j s that lead to different degrees of borrowing: (II): Here, Half-N ( A ) represents a half-normal distribution with scale parameter A , which belongs to the half-t prior family discussed by Gelman [36].”
BackgroundInfluentialA Rose by Another Name? Odor Misnaming is Associated with Linguistic Properties
Hörberg, T., Kurfalı, M., Larsson, M. U., Laukka, E. J., Herman, P. A., & Olofsson, J. (2024) · Cognitive Science · doi:10.1111/cogs.70003 · Cited by 18
Bayesian Astrostatistics: A Backward Look to the Future
Loredo, T. J. (2012) · Springer series in astrostatistics · doi:10.1007/978-1-4614-3508-2_2 · Cited by 42
“…non-nrmalized) priors that are safe to use in simple models (because the likelihood makes the posterior proper) can be dangerous in MLMs, producing improper posteriors; prope but “vague” default priors may hide the problem without truly ameliorating it (Hadjicostas & Berry 1999; Gelman 2006).”
BackgroundFujimoto, M., Nishihara, G. N., Prathep, A., & Terada, R. (2014) · Journal of Applied Phycology · doi:10.1007/s10811-014-0409-0 · Cited by 15
“Weakly informative normal priors were placed on all of the parameters of the model, and a half-cauchy prior was placed on the scale parameter of the models (Gelman 2004, 2006).”
MethodWilson, D. K., Meinzer, C., Lawson, A. B., Alia, K. A., Meyers, D. C., Coulon, S. M., & Lawman, H. G. (2012) · Spatial and Spatio-temporal Epidemiology · doi:10.1016/j.sste.2012.10.001 · Cited by 15
“At the next level of the hierarchy the standard deviations ðr Þ were given non-informative uniform prior distributions on small positive ranges for the regression parameters, individual frailty/uncorrelated heterogeneity (UH), and correlated heterogeneity (CH), respectively, (Gelman, 2006).”
BackgroundFúquene, J. A., Pérez, M.-E., & Pericchi, L. R. (2014) · Brazilian Journal of Probability and Statistics · doi:10.1214/12-bjps207 · Cited by 15
Polson, N. G., & Scott, J. G. (2012) · The Annals of Applied Statistics · doi:10.1214/11-aoas512 · Cited by 35
“Another special case is the half-Cauchy prior on the scale factor λ, studied by Gelman (2006) and Carvalho, Polson and Scott (2010).”
Developing crash modification functions for pedestrian signal improvement
Sacchi, E., Sayed, T., & Osama, A. (2015) · Accident Analysis & Prevention · doi:10.1016/j.aap.2015.07.009 · Cited by 15
Impact of survey design changes on stock assessment advice: sea scallops
Smith, S. J., & Hubley, B. (2013) · ICES Journal of Marine Science · doi:10.1093/icesjms/fst115 · Cited by 27
Liu, H., & Zhu, X. (2016) · ISPRS International Journal of Geo-Information · doi:10.3390/ijgi5070102 · Cited by 15
“(a) History Plot; (b) Posterior density plot; (c) Gelman Rubin statistic plot; (d) Autocorrelation plot; (e) Monte Carlo (MC) error statistics for the regression parameter 𝛽3.”
MethodInfluentialBayesian prospective detection of small area health anomalies using Kullback–Leibler divergence
Rotejanaprasert, C., & Lawson, A. B. (2016) · Statistical Methods in Medical Research · doi:10.1177/0962280216652156 · Cited by 15
Saving My Life: Dynamics of Peer and Staff Corrections Among Therapeutic Community Residents
Doogan, N., & Warren, K. L. (2017) · Substance Use & Misuse · doi:10.1080/10826084.2017.1284236 · Cited by 16
“We used a uniform prior for the random effect variances (Gelman, 2006).”
Thorson, J. T., & Cope, J. M. (2017) · Fisheries Research · doi:10.1016/j.fishres.2017.06.007 · Cited by 15
Chen, N., Carlin, B. P., & Hobbs, B. P. (2018) · Computational Statistics & Data Analysis · doi:10.1016/j.csda.2018.05.002 · Cited by 16
“Other authors have considered hyperprior specifications for Bayesian hierarchical models (Daniels, 1999; Natarajan and Kass, 2000; Spiegelhalter, 2001; Gelman, 2006; Browne and Draper, 2006; Kass and Natarajan, 2006).”
BackgroundTrouvé, R., Bontemps, J., Collet, C., Seynave, I., & Lebourgeois, F. (2018) · Forest Ecology and Management · doi:10.1016/j.foreco.2018.11.007 · Cited by 18
Thomas, F., Yen, J. D. L., & Vesk, P. A. (2019) · Ecology and Evolution · doi:10.1002/ece3.4693 · Cited by 14
“Standard deviations for the observation error and random effects were modeled as posi‐ tive half‐normal distributions, with prior mean 0 and prior stand‐ ard deviation 2 (Gelman, 2006).”
MethodX vs. Y: an analysis of intergenerational differences in transport mode use among young adults
Krueger, R., Rashidi, T. H., & Vij, A. (2019) · Transportation · doi:10.1007/s11116-019-10009-7 · Cited by 23
“Following the examples of Gelman (2006) and Polson and Scott (2012), we also explored the use of half-Cauchy priors.”
Vanhatalo, J., Hartmann, M., & Veneranta, L. (2019) · Bayesian Analysis · doi:10.1214/19-ba1158 · Cited by 15
“In order to alleviate these identifibiality challenges we used weakly informative priors for the variance and length-scale parameters defined through the principles proposed by Gelman (2006), Simpson et al. (2017) and Fuglstad et al. (2018).”
MethodThe implications of unconfounding multisource performance ratings.
Jackson, D. J. R., Michaelides, G., Dewberry, C., Schwencke, B., & Toms, S. (2019) · Journal of Applied Psychology · doi:10.1037/apl0000434 · Cited by 19
Simpson, C. R. (2019) · Social Networks · doi:10.1016/j.socnet.2019.10.001 · Cited by 13
GOVERNMENT SPENDING MULTIPLIERS IN GOOD TIMES AND BAD TIMES: THE CASE OF EMERGING MARKETS
Gomes, F. A. R., Sakurai, S. N., & Soave, G. P. (2020) · Macroeconomic Dynamics · doi:10.1017/s136510052000036x · Cited by 15
“The parameters controlling for the degree of pooling in priors (5) and (7) are set by following Gelman (2006).”
BackgroundJiang, L., Greenwood, C. M. T., Yao, W., et al. (2020) · Scientific Reports · doi:10.1038/s41598-020-66466-z · Cited by 15
“where U is the function defined in (11).”
BackgroundA Bernoulli-Gamma hierarchical Bayesian model for daily rainfall forecasts
Lima, C., Kwon, H., & Kim, Y. (2021) · Journal of Hydrology · doi:10.1016/j.jhydrol.2021.126317 · Cited by 16
Spatiotemporal clustering using Gaussian processes embedded in a mixture model
Vanhatalo, J. P., Foster, S. D., & Hosack, G. R. (2021) · Environmetrics · doi:10.1002/env.2681 · Cited by 15
“Hence, we use a weakly informative half Student-t prior for the SD of the spatial random effects 𝜎𝜙,k (Gelman, 2006) with four degrees of freedom and scale 0.1.”
MethodInfluentialvan Erp, S., & Browne, W. J. (2021) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705511.2021.1915146 · Cited by 22
“The sensitivity of the inverse-Gamma priors to the hyperparameter choice arises especially in situations in which the random effects variance is close to zero (Gelman, 2006).”
BackgroundMethodInfluentialA Bayesian semi-parametric model for thermal proteome profiling
Fang, S., Kirk, P., Bantscheff, M., Lilley, K. S., & Crook, O. M. (2021) · Communications Biology · doi:10.1038/s42003-021-02306-8 · Cited by 18
Paucar-Quishpe, V., Ron‐Román, J., Benítez‐Ortiz, W., Celi-Erazo, M., Berkvens, D. L., Saegerman, C., & Ron-Garrido, L. (2021) · Microorganisms · doi:10.3390/microorganisms9091815 · Cited by 16
Tong, Q., & Gernay, T. (2021) · Natural Hazards · doi:10.1007/s11069-021-05109-6 · Cited by 15
“The precisions and k are assumed to follow an uninformative Half-Cauchy prior (Gelman 2006) with parameters (0, 10).”
BackgroundEstimating global species richness using symbolic data meta‐analysis
Lin, H., Caley, M. J., & Sisson, S. A. (2022) · Ecography · doi:10.1111/ecog.05617 · Cited by 25
Bååth, R. A. (2015) · Behavior Research Methods · doi:10.3758/s13428-015-0591-2 · Cited by 14
“As proposed by Gelman, 2006, the SD parameters sσ and σσ are given half-Cauchy priors with parameter ss,σ and sσ,µ respectively.”
MethodBayesian longitudinal item response modeling with restricted covariance pattern structures
Azevedo, C., Fox, J., & de Andrade, D. F. (2014) · Statistics and Computing · doi:10.1007/s11222-014-9518-5 · Cited by 18
Bayesian hierarchical modeling of cardiac response to particulate matter exposure
McBride, S. J., Norris, G. A., Williams, R. W., & Neas, L. (2009) · Journal of Exposure Science & Environmental Epidemiology · doi:10.1038/jes.2009.58 · Cited by 15
Oguro, M., & Sakai, S. (2015) · American Journal of Botany · doi:10.3732/ajb.1400233 · Cited by 17
Evaluating δ15N–body size relationships across taxonomic levels using hierarchical models
Reum, J. C. P., & Marshall, K. N. (2013) · Oecologia · doi:10.1007/s00442-013-2715-7 · Cited by 15
“We used a uniform [0,10] prior for the standard deviation of each random-effect, and a gamma prior for residual variance, with shape and scale parameters equal to 0.001 (Gelman 2006).”
BackgroundMethodInfluentialAllepuz, A., Sáez, M., Solymosi, N., et al. (2009) · Preventive Veterinary Medicine · doi:10.1016/j.prevetmed.2009.06.005 · Cited by 15
Rúa, S. M. H., Mazumdar, M., & Strawderman, R. L. (2015) · Statistics in Medicine · doi:10.1002/sim.6631 · Cited by 18
Derivation of the Shrinkage Estimates of Quantitative Trait Locus Effects
Xu, S. (2007) · Genetics · doi:10.1534/genetics.107.077487 · Cited by 14
“Exponential distribution (Tibshirani 1996) and half tdistribution (Gelman 2006) have also been used.”
MethodFoulley, J., & Jaffrézic, F. (2009) · Computer Methods and Programs in Biomedicine · doi:10.1016/j.cmpb.2009.05.004 · Cited by 15
Kwok, H., & Lewis, R. J. (2011) · Circulation Cardiovascular Quality and Outcomes · doi:10.1161/circoutcomes.111.960724 · Cited by 16
A Bayesian joint model of menstrual cycle length and fecundity
Lum, K. J., Sundaram, R., Louis, G. M. B., & Louis, T. A. (2015) · Biometrics · doi:10.1111/biom.12379 · Cited by 14
“For example, in (3) we use a uniform prior for the standard deviation of W (σW ) rather than an inverse Gamma, to better control prior mass (Gelman, 2006).”
MethodBayesian Model Selection in Factor Analytic Models
Ghosh, J., & Dunson, D. B. (2008) · Lecture notes in statistics · doi:10.1007/978-0-387-76721-5_7 · Cited by 14
“Ghosh and Dunson (2007) address the above problems by generalizing the idea of Gelman (2006) to induce a new class of robust priors for the factor loadings.”
BackgroundBayesian inference of multi-stage reliability for degradation systems with calibrations
Kong, D., & Cui, L. (2015) · Proceedings of the Institution of Mechanical Engineers Part O Journal of Risk and Reliability · doi:10.1177/1748006x15580752 · Cited by 15
Iijima, H., & Nagaike, T. (2017) · Journal of Forest Research · doi:10.1080/13416979.2017.1305262 · Cited by 15
“U[0, 100], Gelman 2006).”
Chen, Q., Williams, S. Z., Liu, Y., Chihuri, S. T., & Li, G. (2018) · Accident Analysis & Prevention · doi:10.1016/j.aap.2018.08.021 · Cited by 17
Semi-supervised nonparametric Bayesian modelling of spatial proteomics
Crook, O. M., Lilley, K. S., Gatto, L., & Kirk, P. (2022) · The Annals of Applied Statistics · doi:10.1214/22-aoas1603 · Cited by 13
Dynamic Quantile Linear Models: A Bayesian Approach
Gonçalves, K. C. M., Migon, H. S., & Bastos, L. S. (2019) · Bayesian Analysis · doi:10.1214/19-ba1156 · Cited by 15
“In both cases, for the variance of the states W , we considered a half-Cauchy prior distribution, as discussed by Gelman (2006), with scale 25, set as a weakly informa- tive prior distribution.”
MethodJiao, F., Tu, W., Jiménez, S., et al. (2019) · Journal of Biopharmaceutical Statistics · doi:10.1080/10543406.2019.1657132 · Cited by 15
“We discuss the following MAP-Prior methods in detail: the choice of hyper-parameter priors (Gelman 2006), prior robustification (Schmidli, Gsteiger et al. 2014) as a strategy to hedge against heterogeneity, and effective sample size (ESS) (Morita et al. 2008) as a way to measure the strength of…”
MethodInfluentialBorrowing from supplemental sources to estimate causal effects from a primary data source
Boatman, J. A., Vock, D. M., & Koopmeiners, J. S. (2021) · Statistics in Medicine · doi:10.1002/sim.9114 · Cited by 15
“ωq = p ( Ωq|,) = p (|,Ωq) p (Ωq|) ∑2H w=1p (|,Ωw) p (Ωw|) , (5)”
Bayesian Analysis of Extended Cox Model with Time-Varying Covariates Using Bootstrap Prior
Olaniran, O. R., & Abdullah, M. A. A. (2020) · Journal of Modern Applied Statistical Methods · doi:10.22237/jmasm/1604188980 · Cited by 15
“Once the prior hyperparameters and likelihood are determined, the posterior distribution can be sample via MCMC (Gelman, 2006).”
BackgroundUnidirectional response to bidirectional selection on body size II. Quantitative genetics
Rouzic, A. L., Renneville, C., Millot, A., et al. (2020) · Ecology and Evolution · doi:10.1002/ece3.6783 · Cited by 15
Nonstationary multivariate Gaussian processes for electronic health records
Meng, R., Soper, B., Lee, H. K. H., Liu, V. X., Greene, J., & Ray, P. (2021) · Journal of Biomedical Informatics · doi:10.1016/j.jbi.2021.103698 · Cited by 17
A Hierarchical Bayesian Approach for Aerosol Retrieval Using MISR Data
Wang, Y., Jiang, X., Yu, B., & Jiang, M. F. (2013) · Journal of the American Statistical Association · doi:10.1080/01621459.2013.796834 · Cited by 17
“(4) For the above prior to work well, the number of groups, namely P, is to be larger than 5 (Gelman 2006).”
BackgroundAn Empirical Investigation of Bayesian Hierarchical Modeling with Unidimensional IRT Models
Sheng, Y. (2013) · Behaviormetrika · doi:10.2333/bhmk.40.19 · Cited by 16
Korman, J., Decker, A. S., Mossop, B., et al. (2010) · North American Journal of Fisheries Management · doi:10.1577/m09-159.1 · Cited by 15
Ballinger, A., Nally, R. M., & Lake, P. S. (2009) · River Research and Applications · doi:10.1002/rra.1254 · Cited by 15
“Priors for standard deviations were uniform over 0.1– 20 (Gelman, 2006).”
BackgroundInfluentialToxicity Profiling of Engineered Nanomaterials via Multivariate Dose Response Surface ModelingPreprint
Patel, T., Telesca, D., George, S., & Nel, A. E. (2011) · Collection of Biostatistics Research Archive · Cited by 14
Burger, D. A., Schall, R., Ferreira, J. T., & Chen, D. (2020) · Statistics in Medicine · doi:10.1002/sim.8475 · Cited by 14
Unexpected food web responses to low dissolved oxygen in an estuarine fjord
Moriarty, P. E., Essington, T. E., Horne, J. K., Keister, J. E., Li, L., Parker‐Stetter, S. L., & Satô, M. (2020) · Ecological Applications · doi:10.1002/eap.2204 · Cited by 13
Semiparametric Functional Factor Models with Bayesian Rank Selection
Kowal, D. R., & Canale, A. (2023) · Bayesian Analysis · doi:10.1214/23-ba1410 · Cited by 14
Roshan, R., Jahageerdar, S., Jayaraman, J., Chanu, T. I., Bangera, R., & Gilmour, A. R. (2022) · Heredity · doi:10.1038/s41437-022-00509-1 · Cited by 18
“Gelman (2006) noted the popularity of the inverse gamma family of priors is due to its clean mathematical properties, wherein he discourages the use of inverse gamma prior due to the sensitivity of its parameterization towards the posterior inferences.”
BackgroundGrantham, K. L., Kasza, J., Heritier, S. R., Carlin, J. B., & Forbes, A. B. (2022) · BMC Medical Research Methodology · doi:10.1186/s12874-022-01550-8 · Cited by 20
“The half-Cauchy prior has been recommended for variance components in hierarchical models, though more attention has been paid to the cluster variance than to the error variance [10, 37].”
Bayesian multi-level calibration of a process-based maize phenology model
Viswanathan, M., Scheidegger, A., Streck, T., Gayler, S., & Weber, T. K. D. (2022) · Ecological Modelling · doi:10.1016/j.ecolmodel.2022.110154 · Cited by 14
Karpychev, V., Bolgina, T., Александровна, М. С., Zinchenko, V., Ushakov, V., Александрович, И. Г., & Dragoy, O. (2022) · PLoS ONE · doi:10.1371/journal.pone.0276721 · Cited by 15
“For analyses with multiple testing, we adjusted this distribution by decreasing a scale value proportionally to the number of tests [58, 59].”
MethodUncertainty shocks in emerging economies: A global to local approach for identification
Miescu, M. (2023) · European Economic Review · doi:10.1016/j.euroecorev.2023.104437 · Cited by 34
Forecasting Ukrainian Refugee Flows With Organic Data Sources
Wycoff, N., Arab, A., Donato, K. M., Singh, L., Kawintiranon, K., Liu, Y., & Jacobs, E. (2023) · International Migration Review · doi:10.1177/01979183231203931 · Cited by 20
Combining machine learning and econometrics: Application to commercial real estate prices
Francke, M., & de Minne, A. V. (2024) · Real Estate Economics · doi:10.1111/1540-6229.12483 · Cited by 21
Shahriari, S., Sisson, S. A., & Rashidi, T. H. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.126977 · Cited by 16
Kazembe, L. N., & Kandala, N. (2015) · Spatial and Spatio-temporal Epidemiology · doi:10.1016/j.sste.2015.02.001 · Cited by 14
Jointly modeling time-to-event and longitudinal data: a Bayesian approach
Huang, Y., Hu, X. J., & Dagne, G. A. (2013) · Statistical Methods & Applications · doi:10.1007/s10260-013-0242-7 · Cited by 14
The Multigroup Multilevel Categorical Latent Growth Curve Models
Hung, L.-F. (2010) · Multivariate Behavioral Research · doi:10.1080/00273171003680336 · Cited by 15
“A formal discussion on the selection of priors can be found in Congdon (2003), DiCiccio, Kass, Raftery, and Wasserman (1997), Gelman (2006), Kass and Wasserman (1996), and Zhang and Nesselroade (2007).”
BackgroundSemiparametric Regression Analysis via Infer.NET
Luts, J., Wang, S. S. J., Ormerod, J. T., & WAND, M. P. (2018) · Journal of Statistical Software · doi:10.18637/jss.v087.i02 · Cited by 18
Josefsson, M., de Luna, X., Daniels, M. J., & Nyberg, L. H. (2015) · Journal of the Royal Statistical Society Series C (Applied Statistics) · doi:10.1111/rssc.12110 · Cited by 16
“However, such priors are not a good choice for random-effects variances (see, for example, Gelman (2006)); as such, we use the shrinkage priors and half-normal prior for these variances.”
MethodInference in OSNs via Lightweight Partial Crawls
Avrachenkov, K. E., Ribeiro, B., & Sreedharan, J. K. (2016) · doi:10.1145/2896377.2901477 · Cited by 14
“The inverse-gamma is a non-informative conjugate prior if the variance of the estimator is not too small [13], which is generally the case in our application.”
BackgroundA Bayesian inference for time series via copula-based Markov chain models
Sun, L., Lee, C.-S., & Emura, T. (2018) · Communications in Statistics - Simulation and Computation · doi:10.1080/03610918.2018.1529241 · Cited by 14
Impact of using the ultrahigh-energy cosmic ray arrival energies to constrain source associations
Capel, F., & Mortlock, D. J. (2019) · Monthly Notices of the Royal Astronomical Society · doi:10.1093/mnras/stz081 · Cited by 14
“Motivated by the desire to include our knowledge of UHECR phenomenology into the analysis and avoid giving weight to unphysical regions of the parameter space, we choose weakly informative hyperpriors for the these parameters (Gelman 2006; Simpson et al. 2017; Gelman et al. 2017).”
MethodSpatially varying age–period–cohort analysis with application to US mortality, 2002–2016
Chernyavskiy, P., Little, M. P., & Rosenberg, P. S. (2019) · Biostatistics · doi:10.1093/biostatistics/kxz009 · Cited by 14
“Following Gelman (2006), Gelman and others (2008), we assigned weakly informative prior distributions for elements of σ as half-Normal(0, 1).”
MethodZhang, L., Shaby, B. A., & Wadsworth, J. L. (2020) · Journal of the American Statistical Association · doi:10.1080/01621459.2020.1858838 · Cited by 14
“where halfCauchy(1) refers to the positively truncated standard Cauchy distribution (see Gelman, 2006). We implement this methodology for two different Gaussian scale mixture models: that of Huser and Wadsworth (2019), and that of Huser et al. (2017). The priors for θR are δ ∼ U(0, 1) for the former, and β ∼ halfCauch…”
BackgroundMethodInfluentialA Tutorial in Bayesian Mediation Analysis With Latent Variables
Miočević, M. (2019) · Methodology · doi:10.1027/1614-2241/a000177 · Cited by 19
Fradi, A., Feunteun, Y., Samir, C., Baklouti, M., Bachoc, F., & Loubes, J.-M. (2020) · Information Sciences · doi:10.1016/j.ins.2020.09.027 · Cited by 16
Ruiz-Cooley, R. I., Gerrodette, T., Chivers, S. J., & Danil, K. (2021) · Journal of Animal Ecology · doi:10.1111/1365-2656.13478 · Cited by 14
Sex Differences and Exogenous Estrogen Influence Learning and Brain Responses to Prediction Errors
Joue, G., Chakroun, K., Bayer, J., Gläscher, J. P., Zhang, L., Fuß, J., Hennies, N., & Sommer, T. (2021) · Cerebral Cortex · doi:10.1093/cercor/bhab334 · Cited by 14
Kim, Y., Kwon, H., Lima, C., & Sharma, A. (2021) · Geophysical Research Letters · doi:10.1029/2021gl095729 · Cited by 16
“For the overall variance parameter 2 E , we adopt a weakly informative, conjugate prior distribution as done in other studies (Gelman, 2006; Lima et al., 2018; Lunn et al., 2012):”
BackgroundMethodInfluentialMurphy, S. M., Beausoleil, R. A., Stewart, H., & Cox, J. J. (2022) · Global Ecology and Conservation · doi:10.1016/j.gecco.2022.e02109 · Cited by 19
Nardino, M., Silva, F. F. E., Olivoto, T., et al. (2022) · Scientia Agricola · doi:10.1590/1678-992x-2021-0190 · Cited by 14
“The uniform distributions have been used in the Bayesian analysis for both conceptual and practical reasons (Gelman, 2006).”
MethodBlankers, T., Lübke, A. K., & Hennig, R. M. (2015) · Journal of Evolutionary Biology · doi:10.1111/jeb.12686 · Cited by 13
“For all the models (male and female data), parameter expanded, inverse Wishart priors (Gelman, 2006; Hadfield, 2012) were used.”
MethodBisbing, S. M., Cooper, D. J., D’Amore, D. V., & Marshall, K. N. (2015) · Ecohydrology · doi:10.1002/eco.1640 · Cited by 15
“01) for inverse residual variance, Normal (0, 100) for regression coefficients, and uniform (0, 10) for random effects (Gelman, 2006).”
BackgroundMethodInfluentialA stochastic model for acoustic attenuation
Lacaze, B. (2007) · Waves in Random and Complex Media · doi:10.1080/17455030701247919 · Cited by 14
Bayesian and Frequentist Methodologies for Analyzing Method Comparison Studies With Multiple Methods
Choudhary, P. K., & Yin, K. (2010) · Statistics in Biopharmaceutical Research · doi:10.1198/sbr.2010.08096 · Cited by 14
Toxicity profiling of engineered nanomaterials via multivariate dose-response surface modeling
Patel, T., Telesca, D., George, S., & Nel, A. E. (2012) · The Annals of Applied Statistics · doi:10.1214/12-aoas563 · Cited by 16
“We directly sample components when closedform full conditional distributions are available using a Gibbs sampling algorithm [Geman and Geman (1984), Gelfand and Smith (1990)]; otherwise, we use the Metropolis–Hastings (MH) approach [Metropolis et al.”
Bayesian predictive inference of a finite population proportion under selection bias
Nandram, B., Bhatta, D., Bhadra, D., & Shen, G. (2012) · Statistical Methodology · doi:10.1016/j.stamet.2012.08.003 · Cited by 14
Survey data and Bayesian analysis: a cost-efficient way to estimate customer equity
Karvanen, J., Rantanen, A., & Luoma, L. (2014) · Quantitative Marketing and Economics · doi:10.1007/s11129-014-9148-4 · Cited by 18
“For parameters γ, δ, αp, βp, αq and βq we use weakly informative prior distributions (Gelman, 2006).”
Spatial and temporal variation in harvest probabilities for American black duck
Roy, C., Cumming, S. G., & McIntire, E. J. B. (2015) · Ecology and Evolution · doi:10.1002/ece3.1484 · Cited by 14
“Following Gelman (2006), we used half-Cauchy priors for standard deviation parameters (re,rZ).”
Method
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.