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 1,401–1,500 of 4,089
Bakhshi, A. K., & Ahmed, M. M. (2021) · International Journal of Injury Control and Safety Promotion · doi:10.1080/17457300.2021.1907595 · Cited by 19
“…of the Bayesian approach over frequentist methods in dealing with limited datasets since sampling-based Bayesian methods are better equipped to model data with small samples (Browne & Draper, 2006; Gelman, 2006; Hossain et al., 2019; F. L. Mannering et al., 2016; Muth en & Asparouhov, 2012).”
MethodResultModelling typing disfluencies as finite mixture process
Roeser, J., Mæyer, S. D., Leijten, M., & Waes, L. V. (2021) · Reading and Writing · doi:10.1007/s11145-021-10203-z · Cited by 25
“We used a prior of 2.5 for the variance of the variance 2 u to regulate the random intercepts for participants to be normal distributed around the population average (Gelman, 2006).”
MethodInfluentialZhang, X., & Cheng, C. (2022) · International Journal of Environmental Research and Public Health · doi:10.3390/ijerph19020707 · Cited by 19
“5 risk during 2000–2018, (b0t + vt) indicates pollution risk’s temporal variation, t* is observation period calculating from the mid-study period, b1i is the local temporal trend relative to the overall trend b0, and εit ~ N (0, σε(2)) is the Gaussian noise [37].”
BackgroundObserved expectations, news shocks, and the business cycle
Milani, F., & Rajbhandari, A. (2020) · Research in Economics · doi:10.1016/j.rie.2020.02.003 · Cited by 17
“Gelman (2006) discusses how uniform distributions may be unexpectedly informative, with miscalibration toward positive values, when the standard deviations are close to zero, and how results under the previous inverse gamma prior are usually very sensitive to the choice of .”
BackgroundYang, S., & Roger, J. (2009) · Pharmaceutical Statistics · doi:10.1002/pst.400 · Cited by 17
Comparing models for perfluorooctanoic acid pharmacokinetics using Bayesian analysis
Wambaugh, J. F., Barton, H. A., & Setzer, R. W. (2008) · Journal of Pharmacokinetics and Pharmacodynamics · doi:10.1007/s10928-008-9108-2 · Cited by 18
“In most cases the mean and standard deviation hyper-parameters are drawn from the uniform distribution Uðx; min; max Þ¼ 1 max� min ; min \ x \ max [ 31 ].”
BackgroundProspective analysis of infectious disease surveillance data using syndromic information
Corberán‐Vallet, A., & Lawson, A. B. (2014) · Statistical Methods in Medical Research · doi:10.1177/0962280214527385 · Cited by 17
Partitioning Degrees of Freedom in Hierarchical and Other Richly Parameterized Models
Cui, Y., Hodges, J. S., Kong, X., & Carlin, B. P. (2010) · Technometrics · doi:10.1198/tech.2009.08161 · Cited by 17
“Alternative priors on variances or standard deviations have been proposed; Gelman (2006) has provided an influential critique and some alternatives.”
BackgroundHua, D., Jiao, Y., Neves, R. J., & Jones, J. W. (2015) · Ecology and Evolution · doi:10.1002/ece3.1348 · Cited by 28
“The SJ , Pcl j, and Pcd j represent mean values of population characteristics of Si,j, Pcl i,j, and Pcd i,j and among individuals and were assumed to follow uniform distributions between 0 and 1 because uniform prior distributions for variance outperformed in hierarchical models (Gelman 2006).”
BackgroundBiggeri, A., Catelan, D., & Dreassi, E. (2009) · Spatial and Spatio-temporal Epidemiology · doi:10.1016/j.sste.2009.07.006 · Cited by 17
A simulation study to assess statistical methods for binary repeated measures data
Masaoud, E., & Stryhn, H. (2009) · Preventive Veterinary Medicine · doi:10.1016/j.prevetmed.2009.10.004 · Cited by 18
Forecasting Demand for Fashion Goods:A Hierarchical Bayesian Approach
Yelland, P. M., & Dong, X. (2013) · Springer eBooks · doi:10.1007/978-3-642-39869-8_5 · Cited by 18
“To allow for some variation in demand patterns between products, we adopt a so-called hierarchical or multilevel prior (Gelman 2006, Gelman and Hill 2006) in the model.”
MethodFitting Bayesian item response models in Stata and Stan
Grant, R., Furr, D. C., Carpenter, B., & Gelman, A. (2017) · The Stata Journal Promoting communications on statistics and Stata · doi:10.1177/1536867x1701700206 · Cited by 27
Self-citationUsing simulation to evaluate wildlife survey designs: polar bears and seals in the Chukchi Sea
Conn, P. B., Moreland, E. E., Regehr, E. V., Richmond, E. L., Cameron, M. F., & Boveng, P. L. (2016) · Royal Society Open Science · doi:10.1098/rsos.150561 · Cited by 19
“[54]).”
Quasi-complete separation in random effects of binary response mixed models
Sauter, R., & Held, L. (2016) · Journal of Statistical Computation and Simulation · doi:10.1080/00949655.2015.1129539 · Cited by 17
“The prior distribution f(δ) needs to be proper [24], so the posterior f(θ, δ |y) will also be proper.”
BackgroundLiu, Y., Sun, J., Gou, Y., Sun, X., Zhang, D., & Xue, F. (2020) · International Journal of Environmental Research and Public Health · doi:10.3390/ijerph17030879 · Cited by 21
“The variance parameters σ(2)ε, σ(2)z , and σ(2)m were assigned weekly informative priors [31].”
BackgroundPrior distributions for variance parameters in a sparse‐event meta‐analysis of a few small trials
Pateras, K., Nikolakopoulos, S., & Roes, K. C. B. (2020) · Pharmaceutical Statistics · doi:10.1002/pst.2053 · Cited by 19
Allen, B. C., Shao, K., Hobbie, K. A., et al. (2020) · Environment International · doi:10.1016/j.envint.2020.106111 · Cited by 17
“Following Gelman (2006) the prior for β_sigma was defined to be half-Cauchy with scale equal to 5 (a value somewhat greater than the standard deviation we expected a priori).”
MethodspBayesSurv: Fitting Bayesian Spatial Survival Models Using RPreprint
Zhou, H., Hanson, T., & Zhang, J. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1705.04584 · Cited by 17
“Although the default Γ(0.001, 0.001) prior on τ2 has been tested to perform well across various simulation scenarios (Zhou and Hanson 2017), it still should be used with caution in practice; see Gelman (2006) for general suggestions.”
InfluentialSage, A., Langen, M., & de Minne, A. V. (2019) · SSRN Electronic Journal · doi:10.2139/ssrn.3385502 · Cited by 15
Ideas in Ecology and Evolution
(editor, L. A. (2019) · Ideas in Ecology and Evolution · doi:10.4033/iee · Cited by 17
A Continuous Root Water Uptake Isotope Mixing Model
Fu, H., Neil, E. J., Liu, J., & Si, B. (2024) · Water Resources Research · doi:10.1029/2023wr036852 · Cited by 26
“The advantage of the LKJ distribution is that priors for the variance and correlation coefficient are considered separately and avoids tailing problems from the inverse Wishart distribution (Gelman, 2006), which may provide a more accurate estimation.”
Vercelloni, J., Caley, M. J., Kayal, M., Low‐Choy, S., & Mengersen, K. L. (2014) · PLoS ONE · doi:10.1371/journal.pone.0110968 · Cited by 17
Vanem, E., Huseby, A. B., & Natvig, B. (2013) · Environmental and Ecological Statistics · doi:10.1007/s10651-013-0251-6 · Cited by 16
“However, Gelman (2006) recommended to rather use non-informative priors for the variances in order to avoid constraints on the posteriors, and non-informative priors for the variances were also tried as an alternative, i.e. π(σ 2 ι ) ∝ 1 σ 2 ι .”
MethodSmall Area Estimation of Proportions with Constraint for National Resources Inventory Survey
Wang, X., Berg, E. J., Zhu, Z., Sun, D., & Demuth, G. (2018) · Journal of Agricultural Biological and Environmental Statistics · doi:10.1007/s13253-018-0329-6 · Cited by 15
“As in Gelman (2006) and Polson and Scott (2012), the inverted-beta prior is used as the prior distribution for variance parameters, which is equivalent to the use of half-Cauchy prior for the standard deviation parameter.”
MethodCucchi, K., Heße, F., Kawa, N., Wang, C., & Rubin, Y. N. (2019) · Advances in Water Resources · doi:10.1016/j.advwatres.2019.02.003 · Cited by 21
Schürmann, T., Vogt, J., Christ, O., & Beckerle, P. (2019) · Cognitive Processing · doi:10.1007/s10339-019-00928-9 · Cited by 14
“…plausible predictions by implementing a hierarchical dependency where the mean of a participant’s prior is drawn from an inter-individual distribution governing the spread of the individual means, i.e., a hyperparameter or hyperprior (Farrell and Lewandowsky 2018; Gelman 2006; Kruschke 2015).”
BackgroundImpact of Informative Priors on Model Fit Indices in Bayesian Confirmatory Factor Analysis
Edwards, K., & Konold, T. R. (2022) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705511.2022.2126359 · Cited by 21
Elliott, S. A. M., Acou, A., Beaulaton, L., Guitton, J., Réveillac, É., & Rivot, É. (2022) · Progress In Oceanography · doi:10.1016/j.pocean.2022.102924 · Cited by 25
Wu, D., Goldfeld, K. S., & Petkova, E. (2023) · BMC Medical Research Methodology · doi:10.1186/s12874-022-01813-4 · Cited by 16
Selecting relevant moderators with Bayesian regularized meta‐regression
Lissa, C. J. V., van Erp, S., & Clapper, E. (2023) · Research Synthesis Methods · doi:10.1002/jrsm.1628 · Cited by 23
“26 BRMA uses a prior specifically developed for multi-level heterogeneity parameters 28 : a half-Student's t distribution with large variance, student-t þ 3,0,2 : 5 ð Þ .”
MethodExpert elicitation and data noise learning for material flow analysis using Bayesian inference
Dong, J., Liao, J., Huan, X., & Cooper, D. (2023) · Journal of Industrial Ecology · doi:10.1111/jiec.13399 · Cited by 17
“A common choice of prior for the standard deviation of a Gaussian distribution is the Inverse-Gamma distribution, which enables an analytical evaluation of the posterior when the measured data is linear in the parameters of interest [58].”
BackgroundMurphy, A., Burgess, K., Hall, A. J., Aspe, R., & Swinton, P. A. (2023) · The Journal of Strength and Conditioning Research · doi:10.1519/jsc.0000000000004440 · Cited by 31
Eser, M., Mannhardt, S., Gurbuz, C., Brandão, E., & Marburg, S. (2023) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2023.110780 · Cited by 17
Niemeijer, M., Więcek, W., Fu, S., Huppelschoten, S., Bouwman, P., Baze, A., Parmentier, C., Richert, L., Paules, R. S., Bois, F. Y., et al. (2024) · Environmental Health Perspectives · doi:10.1289/ehp11891 · Cited by 21
Estimating density dependence and latent population trajectories with unknown observation error
Hosack, G. R., Peters, G. W., & Hayes, K. R. (2012) · Methods in Ecology and Evolution · doi:10.1111/j.2041-210x.2012.00218.x · Cited by 16
A rigorous small area modelling-study for the helicobacter pylori epidemiology
Strebel, K., Rolle-Kampczyk, U. E., Richter, M., et al. (2010) · The Science of The Total Environment · doi:10.1016/j.scitotenv.2010.03.045 · Cited by 16
Bayesian estimation of intensity surfaces on the sphere via needlet shrinkage and selection
Scott, J. G. (2011) · Bayesian Analysis · doi:10.1214/11-ba611 · Cited by 16
Zeng, L., & Chen, N. (2014) · IIE Transactions · doi:10.1080/0740817x.2014.892230 · Cited by 16
“Two types of priors have been used in the literature (Gelman, 2006): π1(σ 2) ∝ ( 1 σ 2 )r+1 , π2(σ 2) ∝ IG(ω, ω), where the first is a noninformative prior with user-specified r value.”
Bayesian spatial modeling of disease risk in relation to multivariate environmental risk fields
Kim, J., Lawson, A. B., McDermott, S. W., et al. (2009) · Statistics in Medicine · doi:10.1002/sim.3777 · Cited by 16
The Application of a Bayesian Approach to the Analysis of a Complex, Mechanistically Based Model
Jönsson, F., Jonsson, E. N., Bois, F. Y., et al. (2006) · Journal of Biopharmaceutical Statistics · doi:10.1080/10543400600851898 · Cited by 16
“However, it has recently been suggested (Gelman, 2006) that in cases where there is little evidence of variability in the data, inverse gamma distributions tend to not exhibit the desired flexible properties, leading to a posterior that is in fact dominated by the prior distributions.”
BackgroundMethodA hierarchical Bayesian approach for estimating freshwater mussel growth based on tag-recapture data
Tang, M., Jiao, Y., & Jones, J. W. (2013) · Fisheries Research · doi:10.1016/j.fishres.2013.09.005 · Cited by 16
A Bayesian approach for sample size determination in method comparison studies
Yin, K., Choudhary, P. K., Varghese, D., & Goodman, S. R. (2007) · Statistics in Medicine · doi:10.1002/sim.3124 · Cited by 15
“See [21, 22] for some recent suggestions for priors on variance parameters.”
BackgroundBerhane, K. T., & Molitor, N.-T. (2008) · Biostatistics · doi:10.1093/biostatistics/kxm059 · Cited by 16
“Since it has been suggested in the literature (Gelman, 2006) that this choice of prior may prove to be informative, alternative choices of priors are also considered to check for robustness of the substantive findings.”
BackgroundYang, W., Chen, C., & Tempelman, R. J. (2015) · Genetics Selection Evolution · doi:10.1186/s12711-015-0092-x · Cited by 16
Peltola, T., Marttinen, P., & Vehtari, A. (2012) · PLoS ONE · doi:10.1371/journal.pone.0049445 · Cited by 14
A Bayesian Approach to Measurement Bias in Networking Studies
Zhu, L., Robinson, S. E., & Torenvlied, R. (2014) · The American Review of Public Administration · doi:10.1177/0275074014524299 · Cited by 19
“Because prior specification can affect the results of Bayesian inferences (Gelman, Carlin, Stern, & Rubin, 2003; Gill, 2008), we follow Gelman (2006), Li and Baser (2012), and Treier and Jackman (2008) and specify diffuse priors with zero-mean and large-size variance to all parameters. α β θ µ σ …”
BackgroundManica, M., Cobre, P., Rosà, R., & Caputo, B. (2016) · Pest Management Science · doi:10.1002/ps.4315 · Cited by 21
Adaptive spectral estimation for nonstationary multivariate time series
Zhang, S. (2016) · Computational Statistics & Data Analysis · doi:10.1016/j.csda.2016.05.025 · Cited by 16
Li, J., Jin, M., & Zheng, X. (2016) · International Journal of Environmental Research and Public Health · doi:10.3390/ijerph13080772 · Cited by 16
Hierarchical Modeling with Spatial Data
Wikle, C. K. (2010) · Chapman & Hall/CRC interdisciplinary statistics series/Interdisciplinary statistics · doi:10.1201/9781420072884-c7 · Cited by 16
Trends in socioeconomic inequalities in mortality in small areas of 33 Spanish cities
Marí-Dell’Olmo, M., Gotsens, M., Palència, L., et al. (2016) · BMC Public Health · doi:10.1186/s12889-016-3190-y · Cited by 16
A simulation study of methods for selecting subgroup‐specific doses in phase 1 trials
Morita, S., Thall, P. F., & Takeda, K. (2017) · Pharmaceutical Statistics · doi:10.1002/pst.1797 · Cited by 16
Antonelli, J., Zigler, C. M., & Dominici, F. (2017) · Biostatistics · doi:10.1093/biostatistics/kxx003 · Cited by 15
“For example, Gelman and others (1998) addressed the problem of multiple imputation across multiple surveys under the assumption of multivariate normality. Jackson and others (2009) extended these ideas to more complicated structures in a graphical modeling framework. All of these approaches, however, ignore a crucial…”
BackgroundMethodInfluentialSpatiotemporal Patterns of Ground Monitored PM2.5 Concentrations in China in Recent Years
Li, J., Han, X., Li, X., Yang, J., & Li, X. (2018) · International Journal of Environmental Research and Public Health · doi:10.3390/ijerph15010114 · Cited by 17
Rodrigues, H. G., Cornette, R., Clavel, J., Cassini, G. H., Bhullar, B.-A. S., Fernández‐Monescillo, M., Moreno, K., Herrel, A., & Billet, G. (2018) · Royal Society Open Science · doi:10.1098/rsos.171816 · Cited by 16
“The evolutionary rates and measurement error parameters estimated by the model were assigned a weakly informative half-Cauchy prior distribution with scale 25 [29] and the number of shifts was assigned a Poisson distribution with the expected number of shifts set to log(2), which places 50% prior probability on the hy…”
MethodPredicting Danish residential heating energy use from publicly available building characteristics
Kristensen, M. H., Brun, A., & Petersen, S. (2018) · Energy and Buildings · doi:10.1016/j.enbuild.2018.05.011 · Cited by 18
A Bayesian Analysis of Design Parameters in Survey Data Collection
Schouten, B., Mushkudiani, N., Shlomo, N., Durrant, G. B., Lundquist, P., & Wagner, J. R. (2018) · Journal of Survey Statistics and Methodology · doi:10.1093/jssam/smy012 · Cited by 15
“Inverse Gamma distributions are suggested for random effect variance parameters, see Gelman (2006), as they lead to conditionally conjugate prior-posterior pairs.”
BackgroundWestphal‐Fitch, G., Giustolisi, B., Cecchetto, C., Martin, J. S., & Fitch, W. T. (2018) · Frontiers in Psychology · doi:10.3389/fpsyg.2018.01210 · Cited by 19
“…9 | Article 1210 FIGURE 2 | Examples of stimuli sequences for all three grammars with N = 3. priors were specified for the participant random effect variance component to enhance MCMC mixing properties (Gelman, 2006), and diffuse normal priors were specified for the average participant intercept.”
BackgroundConservation reliance of a threatened snake on rice agriculture
Halstead, B. J., Rose, J. P., Reyes, G. A., Wylie, G. D., & Casazza, M. L. (2019) · Global Ecology and Conservation · doi:10.1016/j.gecco.2019.e00681 · Cited by 20
“We specified priors for this model to be uninformative, with normal(mean¼ 0, SD¼ 1.65) priors on mean intercepts and coefficients and half-Cauchy(scale¼ 1) priors on standard deviations (Gelman, 2006).”
MethodInfluentialDynamic borrowing in the presence of treatment effect heterogeneity
Kotalik, A., Vock, D. M., Donny, E. C., et al. (2019) · Biostatistics · doi:10.1093/biostatistics/kxz066 · Cited by 16
“…through a hierarchical structure on the parameters of interest (Han and others, 2017; Chen and others, 2011; Neuenschwander and others, 2010; Spiegelhalter and others, 2004), but hierarchical models have been shown to be sensitive to the hyperprior on the variance parameters (Gelman, 2006).”
BackgroundDifferent Roles of Prior Distributions in the Single Mediator Model with Latent Variables
Miočević, M., Levy, R., & MacKinnon, D. P. (2020) · Multivariate Behavioral Research · doi:10.1080/00273171.2019.1709405 · Cited by 19
“…an inverse-gamma prior with both hyperparameters close to 0 (e.g., equal to 0.001) for a variance parameter can be considered uninformative in some settings, but as Gelman (2006) showed, this prior is not uninformative for a level-2 variance parameter in a multilevel model with only 8 clusters.”
ResultInfluentialDuffy, B. A., Choy, M., & Lee, J. H. (2020) · Cell Reports · doi:10.1016/j.celrep.2020.01.095 · Cited by 15
“Weakly informative priors were used (Gelman et al., 2008) as these are typically preferred over noninformative priors which can lead to improper posterior distributions (Gelman, 2006).”
Reparameterized inverse gamma regression models with varying precision
Bourguignon, M., & Gallardo, D. I. (2020) · Statistica Neerlandica · doi:10.1111/stan.12221 · Cited by 16
“Gelman (2006) studied the IG distribution as a prior distributions for variance parameters in hierarchical models.”
BackgroundMethodSato, Y., Kozono, J., Nishihara, G. N., et al. (2020) · Phycologia · doi:10.1080/00318884.2020.1777384 · Cited by 16
“…chains were sampled to provide at least 1000 samples for each parameter, and convergence was assessed visually and by estimating R. Informative normally distributed priors were placed on the parameter means and half-Cauchy priors on the standard deviations, as recommended by Gelman (2004, 2006).”
MethodEnvironmental Change, If Unaccounted, Prevents Detection of Cryptic Evolution in a Wild Population
Potter, T., Bassar, R. D., Bentzen, P., Ruell, E. W., Torres‐Dowdall, J., Handelsman, C. A., Ghalambor, C. K., Travis, J., Reznick, D. N., & Coulson, T. N. (2020) · The American Naturalist · doi:10.1086/711874 · Cited by 20
Zhang, L., Shaby, B. A., & Wadsworth, J. L. (2021) · Lancaster EPrints (Lancaster University) · doi:10.6084/m9.figshare.13342224.v2 · Cited by 16
Batley, P. N., & Hedges, L. V. (2021) · Behavior Research Methods · doi:10.3758/s13428-020-01522-0 · Cited by 19
“This allows the parameters of the priors to be estimated from the data rather than specifying them to have subjective information, especially for small sample data like those in the present study (Efron & Morris, 1975; Gelman, 2006; James & Stein, 1960).”
MethodMultilevel Time Series Modelling of Mobility Trends in the Netherlands for Small Domains
Boonstra, H. J., van den Brakel, J., & Das, S. (2021) · Journal of the Royal Statistical Society Series A (Statistics in Society) · doi:10.1111/rssa.12700 · Cited by 16
“In the case of a diagonal matrix with equal or unequal diagonal elements, half- Cauchy priors are used for the standard deviations in V. Gelman (2006) demonstrates that these priors are better default priors than the more common inverse gamma priors for the variances.”
MethodGeneralized Topp-Leone-Weibull AFT Modelling: A Bayesian Analysis with MCMC Tools using R and Stan
Ashraf-Ul-Alam, M., & Khan, A. A. (2021) · Austrian Journal of Statistics · doi:10.17713/ajs.v50i5.1166 · Cited by 16
Multivariate Bayesian spatio-temporal P-spline models to analyze crimes against women
Fuenzalida, G. V., Goicoa, T., & Ugarte, M. D. (2021) · Biostatistics · doi:10.1093/biostatistics/kxab042 · Cited by 19
Peterson, E. N., Chou, D., Moller, A., Gemmill, A., Say, L., & Alkema, L. (2022) · Statistics in Medicine · doi:10.1002/sim.9335 · Cited by 16
Investigating the role of working memory resources across aesthetic and non-aesthetic judgements
Bara, I., Binney, R. J., & Ramsey, R. (2022) · Quarterly Journal of Experimental Psychology · doi:10.1177/17470218221101876 · Cited by 20
“Also, by using weakly informative priors, we allow for the possibility of large effects, should they exist in the data (Gelman, 2006; Gelman et al., 2013; Gelman & Hill, 2007; Lemoine, 2019).”
Olivares, J., Moraux, E., Sarro, L. M., et al. (2017) · Astronomy and Astrophysics · doi:10.1051/0004-6361/201731996 · Cited by 16
“These priors fall in the category of weakly informative ones (see Gelman 2006).”
BackgroundFully Bayesian logistic regression with hyper-LASSO priors for high-dimensional feature selection
Li, L., & Yao, W. (2018) · Journal of Statistical Computation and Simulation · doi:10.1080/00949655.2018.1490418 · Cited by 16
“Such alternative penalties or priors to Laplace include, but not limited to: SCAD (Fan and Li, 2001), horseshoe and normal/inverted-beta priors (Gelman, 2006; Carvalho et al., 2009, 2010; Polson and Scott, 2012c; van der Pas et al., 2014), NEG priors (Griffin and Brown, 2011), Normal/Gamma (Caron and Doucet, 2008; Gri…”
BackgroundEffects of Multiple Levels of Social Organization on Survival and Abundance
Ward, E. J., Semmens, B. X., Holmes, E. E., & Balcomb, K. C. (2010) · Conservation Biology · doi:10.1111/j.1523-1739.2010.01600.x · Cited by 16
“Following Gelman (2006), we used uniform (0, 10) priors on standard deviations of random effects for social groupings and temporal deviations.”
MethodHe, Y., Landrum, M. B., & Zaslavsky, A. M. (2014) · Statistics in Medicine · doi:10.1002/sim.6173 · Cited by 16
Osada, Y., Kuriyama, T., Asada, M., Yokomizo, H., & Miyashita, T. (2015) · Population Ecology · doi:10.1007/s10144-015-0498-x · Cited by 16
“The standard deviation of the random effect followed a half-Cauchy distribution, which was proposed as a weak informative prior by Gelman (2006).”
BackgroundMörk, A.-K., Jönsson, F., & Johanson, G. (2009) · Toxicology and Applied Pharmacology · doi:10.1016/j.taap.2009.07.033 · Cited by 16
Bayesian point event modeling in spatial and environmental epidemiology
Lawson, A. B. (2012) · Statistical Methods in Medical Research · doi:10.1177/0962280212446328 · Cited by 19
Bayesian Spatial Modeling of Data from Unit‐Count Surveys of Fish in Streams
Webster, R. A., Pollock, K. H., Ghosh, S. K., et al. (2008) · Transactions of the American Fisheries Society · doi:10.1577/t06-138.1 · Cited by 16
“We used Uniform(0,20) priors for σ Φ in the CAR model and σ τ in the spatial independence model, which follows the suggestion of Gelman (2006 ) to use a uniform density with a large upper bound (relative to the posterior density) as a noninformative proper prior for such parameters.”
MethodOuyang, B., Sinha, D., Slate, E. H., & Bakel, A. B. V. (2012) · Statistics in Medicine · doi:10.1002/sim.5717 · Cited by 14
Finucane, M. M., Paciorek, C. J., Stevens, G. A., & Ezzati, M. (2014) · Journal of the American Statistical Association · doi:10.1080/01621459.2014.937487 · Cited by 16
“Whereas Rodriguez, Dunson, and Taylor (2009) set each τ parameter equal to 1, we assign a uniform hyperprior to each of the τ ’s (Gelman 2006), truncating such that τ > 0.00001 (as smaller values are not identifiable from one another).”
BackgroundMethodInfluentialTerada, R., Vo, T. D., Nishihara, G. N., Matsumoto, K., Kokubu, S., Watanabe, Y., & Kawaguchi, S. (2016) · Phycological Research · doi:10.1111/pre.12143 · Cited by 16
“…was assessed visually and by estimating R. Informative normal priors were placed on all of the parameters of the model by using values from previous studies (Kokubu et al. 2015; Terada et al. 2016a), and a half-cauchy prior was placed on the scale parameter of the models (Gelman 2004,2006).”
MethodUsing phase II data for the analysis of phase III studies: An application in rare diseases
Wandel, S., Neuenschwander, B. E., Röver, C., & Friede, T. (2017) · Clinical Trials · doi:10.1177/1740774517699409 · Cited by 18
Bayesian Small Area Estimation for Skewed Business Survey Variables
Fabrizi, E., Ferrante, M. R., & Trivisano, C. (2017) · Journal of the Royal Statistical Society Series C (Applied Statistics) · doi:10.1111/rssc.12254 · Cited by 16
“According to result (11), the parameter b0 can be interpreted in terms of degrees of freedom of the marginal prior p.τ /, so the choice of b0 =2 is in line with selecting half-t-priors with a very small number of degrees of freedom (Gelman, 2006).”
BackgroundMethodUsing accelerated drug stability results to inform long‐term studies in shelf life determination
Faya, P., Seaman, J. W., & Stamey, J. D. (2018) · Statistics in Medicine · doi:10.1002/sim.7663 · Cited by 16
Borlongan, I. A., Maeno, Y., Kozono, J., et al. (2019) · Phycologia · doi:10.1080/00318884.2019.1571355 · Cited by 16
“Appropriate prior probability distributions were assigned to all parameters of the model, and a half-Cauchy prior was placed on the scale parameter of the models (Gelman 2004, 2006).”
MethodThe Horseshoe-Like Regularization for Feature Subset Selection
Bhadra, A., Datta, J., Polson, N. G., & Willard, B. T. (2019) · Sankhya B · doi:10.1007/s13571-019-00217-7 · Cited by 26
Paulewicz, B., & Blaut, A. (2020) · Behavior Research Methods · doi:10.3758/s13428-020-01370-y · Cited by 15
Going beyond mean effect size: Presenting prediction intervals for on-farm network trial analyses
Laurent, A., Miguez, F. E., Kyveryga, P., & Makowski, D. (2020) · European Journal of Agronomy · doi:10.1016/j.eja.2020.126127 · Cited by 21
The genetic architecture of sexual dimorphism in the mossCeratodon purpureus
Kollar, L. M., Kiel, S., James, A. J., Carnley, C. T., Scola, D. N., Clark, T., Khanal, T., Rosenstiel, T. N., Gall, E. T., Grieshop, K., et al. (2021) · Proceedings of the Royal Society B Biological Sciences · doi:10.1098/rspb.2020.2908 · Cited by 19
“To determine the robustness of the posterior distribution to the prior [43,44], we compared models to other priors.”
State space models with spatial deformation
Morales, F. E. C., Gamerman, D., & Paez, M. S. (2012) · Environmental and Ecological Statistics · doi:10.1007/s10651-012-0215-2 · Cited by 16
“Denote by v UG = (v n + 1 , . . . , v n + m ) the vector of variances in the process at ungauged locations.”
BackgroundEvaluation of Bayesian models for focused clustering in health data
Ma, B., Lawson, A. B., & Liu, Y. (2007) · Environmetrics · doi:10.1002/env.850 · Cited by 16
Strategic mobilization: Why proportional representation decreases voter mobilization
Rainey, C. (2014) · Electoral Studies · doi:10.1016/j.electstud.2014.10.008 · Cited by 16
A flexible parametric approach to examining spatial variation in relative survival
Cramb, S. M., Mengersen, K. L., Lambert, P. C., RYAN, L. M., & Baade, P. D. (2016) · Statistics in Medicine · doi:10.1002/sim.7071 · Cited by 16
Forecasting sub-national trends in COVID-19 vaccine uptake in the UKPreprint
de Figueiredo, A. (2020) · medRxiv · doi:10.1101/2020.12.17.20248382 · Cited by 16
Koch, B., Silvestro, D., & Foster, J. G. (2020) · SocArXiv (OSF Preprints) · doi:10.31235/osf.io/659bt · Cited by 17
Bayesian fusion of multispectral and hyperspectral images using a block coordinate descent method
Wei, Q., Dobigeon, N., & Tourneret, J. (2015) · Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing · doi:10.1109/whispers.2015.8075373 · Cited by 19
Ferrea, E., Franke, J., Morel, P., & Gail, A. (2022) · Scientific Reports · doi:10.1038/s41598-022-13866-y · Cited by 17
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.