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 801–900 of 4,089
Buoro, M., Giménez, O., & Prévost, É. (2011) · Evolution · doi:10.1111/j.1558-5646.2011.01484.x · Cited by 40
Hennessey, V., Rosner, G. L., Bast, R. C., & Chen, M. (2010) · Biometrics · doi:10.1111/j.1541-0420.2010.01403.x · Cited by 38
“Gelman (2006) proposed the use of half-Cauchy prior distributions for the standard deviations (σ) of second-level parameters when the number of groups is small.”
BackgroundMethodFalcy, M. (2015) · Journal of Animal Ecology · doi:10.1111/1365-2656.12297 · Cited by 29
Burke, D. L., Bujkiewicz, S., & Riley, R. D. (2016) · Statistical Methods in Medical Research · doi:10.1177/0962280216631361 · Cited by 36
“and Gelman.(26) These refer to distributions that are set up so that the information they provide is weak but contain only realistic values for the variance.”
BackgroundMethodResultInfluentialBody density and diving gas volume of the northern bottlenose whale (Hyperoodon ampullatus)
Miller, P. J. O., Narazaki, T., Isojunno, S., Aoki, K., Smout, S., & Sato, K. (2016) · Journal of Experimental Biology · doi:10.1242/jeb.137349 · Cited by 49
Probabilistic mortality forecasting with varying age-specific survival improvements
Bohk-Ewald, C., & Rau, R. (2017) · Genus · doi:10.1186/s41118-016-0017-8 · Cited by 39
“A hierarchical model can combine within-group and between-group information automatically (Carlin and Louis 2008; Gelman 2006; Gelman et al. 2003; Jackman 2009; Kruschke 2011).”
MethodBayesian Models for Multiple Outcomes Nested in Domains
Thurston, S. W., Ruppert, D., & Davidson, P. W. (2009) · Biometrics · doi:10.1111/j.1541-0420.2009.01224.x · Cited by 36
“Our diagnostics for multiple chains included (a) visual examination of all traceplots, (b) the Gelman-Rubin diagnostic R̂ [11], and (c) mcgibbsit [28]; and for single chains the Raftery–Lewis diagnostic [18].”
MethodGuidance on the implementation and reporting of a drug safety Bayesian network meta‐analysis
Ohlssen, D., Price, K. L., Amy Xia, H., Hong, H., Kerman, J., Fu, H., Quartey, G., Heilmann, C. R., Ma, H., & Carlin, B. P. (2014) · Pharmaceutical Statistics · doi:10.1002/pst.1592 · Cited by 34
“as half-Cauchy (or half-normal) with an appropriate scale parameter has been recommended [40].”
BackgroundEstimating exposure response functions using ambient pollution concentrations
Shaddick, G., Lee, D., Zidek, J. V., & Salway, R. (2008) · The Annals of Applied Statistics · doi:10.1214/08-aoas177 · Cited by 34
Bandyopadhyay, D., Lachos, V. H., Abanto‐Valle, C. A., & Ghosh, P. (2010) · Statistics in Medicine · doi:10.1002/sim.4031 · Cited by 39
The evolution of anthropoid molar proportions
Carter, K. E., & Worthington, S. (2016) · BMC Evolutionary Biology · doi:10.1186/s12862-016-0673-5 · Cited by 41
“It has been suggested that parameter expanded priors are preferable when weakly informative priors are sought [102].”
BackgroundGrund, S., Lüdtke, O., & Robitzsch, A. (2017) · Journal of Educational and Behavioral Statistics · doi:10.3102/1076998617738087 · Cited by 41
Behmanesh, I., Yousefianmoghadam, S., Nozari, A., et al. (2018) · Mechanical Systems and Signal Processing · doi:10.1016/j.ymssp.2018.01.033 · Cited by 34
The ecology of outdoor rape: The case of Stockholm, Sweden
Ceccato, V. A., Li, G., & Haining, R. (2018) · European Journal of Criminology · doi:10.1177/1477370818770842 · Cited by 38
“To σ v and σu , the standard deviations of the two sets of random effects, a normal prior N(0,1) is assigned; to ensure that a standard deviation parameter is non-negative, this normal prior is truncated so that it takes only non-negative values (Gelman, 2006).”
BackgroundTopic Modelling for Medical Prescription Fraud and Abuse Detection
Zafari, B., & Ekin, T. (2018) · Journal of the Royal Statistical Society Series C (Applied Statistics) · doi:10.1111/rssc.12332 · Cited by 37
“This choice of priors keeps shrinkage close to zero without producing sparsity (Gelman, 2006; Polson and Scott, 2012).”
BackgroundChen, C., Steibel, J. P., & Tempelman, R. J. (2017) · Genetics · doi:10.1534/genetics.117.202259 · Cited by 33
“Furthermore, 185 for all analyses in this paper, we specify Gelman’s non-informative prior (Gelman 2006) for 2 e 186”
MethodMullane, M. J., Barnett, T. C., Cannon, J. W., Carapetis, J. R., Christophers, R., Coffin, J., Jones, M. A., Marsh, J. A., Loughlin, F. M., O’Donnell, V., et al. (2019) · BMJ Open · doi:10.1136/bmjopen-2019-030635 · Cited by 33
“In the event of convergence issues, we will adopt analogous methods such as generalised estimating equations (GEE) that have fewer assumptions compared with GLMMs or Bayesian hierarchical models with sensible priors on variance components.(45) However, GEEs provide only population-level inference and, as per a GLMM, w…”
MethodDegradation modeling with subpopulation heterogeneities based on the inverse Gaussian process
Xu, A., Hu, J., & Wang, P. (2019) · Applied Mathematical Modelling · doi:10.1016/j.apm.2019.12.017 · Cited by 34
Damblin, G., & Gaillard, P. (2020) · Nuclear Engineering and Design · doi:10.1016/j.nucengdes.2019.110391 · Cited by 35
“However, as the posterior distribution of σ2j puts a substential probability mass around 0, a non negligible impact of the value on π(b, σ2|z′,f ,Xf ) may be suspected (Gelman, 2006).”
BackgroundInfluentialMalick, M. J., Hunsicker, M., Haltuch, M., Parker‐Stetter, S. L., Berger, A. M., & Marshall, K. N. (2020) · Marine Ecology Progress Series · doi:10.3354/meps13286 · Cited by 36
MARTICORENA, F. M., CARVALHO, A., OLIVEIRA, L. F. D., DOLAN, E., GUALANO, B., SWINTON, P., & SAUNDERS, B. (2021) · Medicine & Science in Sports & Exercise · doi:10.1249/mss.0000000000002635 · Cited by 34
Zhou, Y., Jiang, X., Fu, C., et al. (2021) · Accident Analysis & Prevention · doi:10.1016/j.aap.2021.106183 · Cited by 34
Wang, X., Suttner, L. H., Jemielita, T., & Li, X. N. (2021) · Journal of Biopharmaceutical Statistics · doi:10.1080/10543406.2021.2011743 · Cited by 48
Stegmueller, D. (2013) · Electoral Studies · doi:10.1016/j.electstud.2013.06.005 · Cited by 60
“I also consider a prior which is not defined for the variance (or precision) but for the standard deviation, as suggested by Gelman (2006).”
BackgroundCausal explanations for the evolution of ‘low gear’ locomotion in insular ruminants
Rozzi, R., Varela, S., Bover, P. R., & Martin, J. M. (2020) · Journal of Biogeography · doi:10.1111/jbi.13942 · Cited by 35
“…effect to avoid singularity, which commonly occurs with small numbers of randomeffect levels (e.g. <5; see Bates, Kliegl, Vasishth, & Baayen, 2018; Gelman, 2006; Matuschek, Kliegl, Vasishth, Baayen, & Bates, 2017) and which we observed in a preliminary analysis (subfamily includes only two…”
BackgroundTerada, R., Takaesu, M., Borlongan, I. A., & Nishihara, G. N. (2021) · Journal of Applied Phycology · doi:10.1007/s10811-021-02439-7 · Cited by 33
“Informative normal priors were placed on all parameters of the model by using values from a previous study (Kokubu et al. 2015), and a half-Cauchy prior distribution was placed on the scale parameter of the models (Gelman 2004, 2006).”
MethodObjective Bayesian Inference for a Generalized Marginal Random Effects Model
Bodnar, O., Link, A., & Elster, C. (2015) · Bayesian Analysis · doi:10.1214/14-ba933 · Cited by 33
Tabandeh, A., & Gardoni, P. (2015) · ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems Part A Civil Engineering · doi:10.1061/ajrua6.0000817 · Cited by 33
Measuring Governance: Implications of Conceptual Choices
Bersch, K., & Botero, S. (2014) · European Journal of Development Research · doi:10.1057/ejdr.2013.49 · Cited by 37
Overstall, A. M., King, R., Bird, S. M., Hutchinson, S. J., & Hay, G. (2013) · Statistics in Medicine · doi:10.1002/sim.6047 · Cited by 42
Probabilistic Modeling of Imaging, Genetics and Diagnosis
Batmanghelich, N. K., Dalca, A., Quon, G., Sabuncu, M., & Golland, P. (2016) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2016.2527784 · Cited by 33
“2) Prior Over Variance of Residual : It is common to assume an uninformative prior distribution2 for the variance of residuals [19].”
BackgroundGenotype × Environment Interaction for Plant Density Response in Maize (Zea mays L.)
Edwards, J. W. (2016) · Crop Science · doi:10.2135/cropsci2015.07.0408 · Cited by 38
Köther, U., Vettorazzi, E., Veckenstedt, R., Hottenrott, B., Bohn, F., Scheu, F., Pfueller, U., Roesch-Ely, D., & Moritz, S. (2017) · Journal of Experimental Psychopathology · doi:10.5127/jep.054516 · Cited by 35
“For both models, we assumed uninformative priors following Gelman (2006).”
MethodAdaptive Bayesian Time–Frequency Analysis of Multivariate Time Series
Li, Z., & Krafty, R. T. (2018) · Journal of the American Statistical Association · doi:10.1080/01621459.2017.1415908 · Cited by 38
“Based on Gelman (2006), we assume priors for λ(2)c jkq, λ(2)b jkq, and λ 2 d j jq to be Uniform(0, κ ), where κ is a known large constant.”
MethodZhang, J., Ko, C., Nie, L., Chen, Y., & Tiwari, R. C. (2018) · Statistical Methods in Medical Research · doi:10.1177/0962280218754928 · Cited by 33
Network‐scale effects of invasive species on spatially‐structured amphibian populations
Manenti, R., Falaschi, M., Monache, D. D., Marta, S., & Ficetola, G. F. (2019) · Ecography · doi:10.1111/ecog.04571 · Cited by 34
“Specifically, for regression coefficients, growth rate and density dependent effects we used a vague normal prior with mean = 0 and variance = 100; for the standard deviations of the random effects we used a halfCauchy prior (Gelman 2006).”
MethodspBayesSurv: Fitting Bayesian Spatial Survival Models Using R
Zhou, H., Hanson, T., & Zhang, J. (2020) · Journal of Statistical Software · doi:10.18637/jss.v092.i09 · Cited by 64
“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.”
InfluentialBiron, M., Peretz, H., & Lupo, K. T. (2020) · Sustainability · doi:10.3390/su12229773 · Cited by 46
Genome‐enabled prediction for sparse testing in multi‐environmental wheat trials
Crespo‐Herrera, L., Howard, R., Piepho, H., Pérez‐Rodríguez, P., Montesinos-López, O. A., Burgueño, J. A., Singh, R. P., Mondal, S., Jarquín, D., & Crossa, J. L. (2021) · The Plant Genome · doi:10.1002/tpg2.20151 · Cited by 40
“Gelman (2006) has argued against these types of priors because they are sensitive to the selection of hyperparameters, which is true for small sample sizes, but it is well known that as the sample size increases, the likelihood dominates the prior.”
BackgroundJacquemin, S., Doll, J. C., Johnson, L. T., & Newell, S. E. (2023) · Harmful Algae · doi:10.1016/j.hal.2023.102374 · Cited by 36
Robertson, D., Chiew, F. H. S., & Potter, N. J. (2023) · Journal of Hydrology · doi:10.1016/j.jhydrol.2023.129322 · Cited by 35
Sinclair, F. H., Stone, G. N., Nicholls, J. A., Cavers, S., Gibbs, M., Butterill, P. T., Wagner, S., Ducousso, A. A., Gerber, S., Petit, R. J., et al. (2015) · Evolutionary Applications · doi:10.1111/eva.12329 · Cited by 40
Ong, M. E. H., Earnest, A., Shahidah, N., Ng, W. M., Foo, C., & Nott, D. J. (2011) · Annals of Emergency Medicine · doi:10.1016/j.annemergmed.2010.12.014 · Cited by 34
Yan, Z., & Haimes, Y. Y. (2010) · Reliability Engineering & System Safety · doi:10.1016/j.ress.2010.02.014 · Cited by 33
Sparse Log Gaussian Processes via MCMC for Spatial Epidemiology
Vanhatalo, J. P., & Vehtari, A. (2007) · Cited by 43
Bayesian methods for latent trait modelling of longitudinal data
Dunson, D. B. (2007) · Statistical Methods in Medical Research · doi:10.1177/0962280206075309 · Cited by 32
Streamlined mean field variational Bayes for longitudinal and multilevel data analysis
Lee, C. Y. Y., & WAND, M. P. (2016) · Biometrical Journal · doi:10.1002/bimj.201500007 · Cited by 32
“C© 2016 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.biometrical-journal.com As explained in Gelman (2006), (14) is an attractive means by which weakly informative priors can be imposed on the σu s. Result 5 of Wand et al. (2011) leads to the following equivalent distributional statement: σ 2u…”
BackgroundBayesian joint modeling of longitudinal and spatial survival AIDS data
Martins, R., Silva, G. L., & Andreozzi, V. (2016) · Statistics in Medicine · doi:10.1002/sim.6937 · Cited by 38
The Scaled Beta2 Distribution as a Robust Prior for Scales
Pérez, M.-E., Pericchi, L. R., & Ramírez, I. C. (2016) · Bayesian Analysis · doi:10.1214/16-ba1015 · Cited by 33
Cooper, K., Alexander, L., Brandie, D., Brown, V. T., Greig, L., Harrison, I., MacLean, C., Mitchell, L., Morrissey, D., Moss, R. A., et al. (2023) · Health Technology Assessment · doi:10.3310/tfws2748 · Cited by 55
Almaraz, P., Green, A. J., Aguilera, E., Rendón, M. Á., & Bustamante, J. (2012) · Journal of Animal Ecology · doi:10.1111/j.1365-2656.2012.01972.x · Cited by 34
Joint Modeling of Multiple Crimes: A Bayesian Spatial Approach
Liu, H., & Zhu, X. (2017) · ISPRS International Journal of Geo-Information · doi:10.3390/ijgi6010016 · Cited by 33
“Regarding the univariate model, we note that standard deviation uniform priors are now commonly used [56].”
MethodMeasuring and modeling salience with the theory of visual attention
Krüger, A., Tünnermann, J., & Scharlau, I. (2017) · Attention Perception & Psychophysics · doi:10.3758/s13414-017-1325-6 · Cited by 35
“We chose uninformative priors except for the weakly informative variance priors favoring variance values close to zero (Gelman, 2006).”
MethodModeling Random Effects Using Global–Local Shrinkage Priors in Small Area Estimation
Tang, X., Ghosh, M., Ha, N. S., et al. (2018) · Journal of the American Statistical Association · doi:10.1080/01621459.2017.1419135 · Cited by 32
Modeling and Simulation of Pretomanid Pharmacokinetics in Pulmonary Tuberculosis Patients
Lyons, M. A. (2018) · Antimicrobial Agents and Chemotherapy · doi:10.1128/aac.02359-17 · Cited by 36
Model-Based County Level Crop Estimates Incorporating Auxiliary Sources of Information
Erciulescu, A. L., Cruze, N. B., & Nandram, B. (2018) · Journal of the Royal Statistical Society Series A (Statistics in Society) · doi:10.1111/rssa.12390 · Cited by 34
“The work of Browne and Draper (2006) motivates the use of a prior uniform distribution for the random-effects variance components; see also the discussion in Gelman (2006).”
MethodKim, T.-J., Kwon, H.-H., & Lima, C. (2018) · Journal of Hydrology · doi:10.1016/j.jhydrol.2018.07.082 · Cited by 32
Variance Prior Forms for High-Dimensional Bayesian Variable Selection
Moran, G. E., Ročková, V., & George, E. I. (2019) · Bayesian Analysis · doi:10.1214/19-ba1149 · Cited by 33
Nations, J. A., Heaney, L. R., Demos, T. C., Achmadi, A. S., Rowe, K. C., & Esselstyn, J. A. (2019) · Biological Journal of the Linnean Society · doi:10.1093/biolinnean/blz085 · Cited by 38
“We fitted regularizing normal priors on the group-level effects [N(0, 1.5)] to prevent Markov chain Monte Carlo (MCMC) chains from searching unreasonable areas of parameter space (Gelman, 2006; McElreath, 2016).”
MethodInfluentialHartmann, M., & Vanhatalo, J. P. (2018) · Statistics and Computing · doi:10.1007/s11222-018-9836-0 · Cited by 44
Reconciling the water balance of large lake systems
Gronewold, A. D., Smith, J. P., Read, L., & Crooks, J. (2020) · Advances in Water Resources · doi:10.1016/j.advwatres.2020.103505 · Cited by 37
Road environments and bicyclist route choice: The cases of Davis and San Francisco, CA
Fitch, D. T., & Handy, S. (2020) · Journal of Transport Geography · doi:10.1016/j.jtrangeo.2020.102705 · Cited by 38
Comparing traditional and Bayesian approaches to ecological meta‐analysis
Pappalardo, P., Ogle, K., Hamman, E. A., Bence, J. R., Hungate, B. A., & Osenberg, C. W. (2020) · Methods in Ecology and Evolution · doi:10.1111/2041-210x.13445 · Cited by 35
“Given that even diffuse priors for 2 among can influence the posterior for 2 among , particularly under small group size (Gelman, 2006), we explored five different priors for 2 among (Figures S12–S15).”
Evaluating the impact of a small number of areas on spatial estimation
Aswi, A., Cramb, S. M., Duncan, E. W., et al. (2020) · International Journal of Health Geographics · doi:10.1186/s12942-020-00233-1 · Cited by 32
“Ideally, trying different distributions (such as a Uniform prior on the standard deviation [27]), as well as considering different hyperparameters within each distribution, is recommended.”
BackgroundUncertainty matters: Bayesian modeling of bicycle crashes with incomplete exposure data
Xu, P., Bai, L., Pei, X., Wong, S. C., & Zhou, H. (2021) · Accident Analysis & Prevention · doi:10.1016/j.aap.2021.106518 · Cited by 33
“…( ε , ε ) prior is sensitive to the value of ε if the true variance is close to zero, a uniform (0,10) was assigned to σ u , s , b , v following Gelman (2006 Once the uncertainty due to incomplete exposure data is adjusted, we can calculate the expected risk of collisions with motor vehicles…”
BackgroundA Bayesian central equivalent dose model for optically stimulated luminescence dating
Combès, B., Lanos, P., Mercier, N., et al. (2015) · HAL (Le Centre pour la Communication Scientifique Directe) · Cited by 32
Kneib, T., Konrath, S., & Fahrmeir, L. (2010) · Journal of the Royal Statistical Society Series C (Applied Statistics) · doi:10.1111/j.1467-9876.2010.00723.x · Cited by 34
“Note, however, that a flat prior p.δl/∝1, which was recommended by Gelman (2006), can be written as an improper inverse gamma prior as well with al =−0:5 and bl =0.”
BackgroundThe Open-Faced Sandwich Adjustment for MCMC Using Estimating Functions
Shaby, B. A. (2013) · Journal of Computational and Graphical Statistics · doi:10.1080/10618600.2013.842174 · Cited by 44
“Weakly informative half-Cauchy priors (see Gelman 2006) were used for both parameters.”
MethodESTIMATING AND VISUALIZING FITNESS SURFACES USING MARK-RECAPTURE DATA
Giménez, O., Grégoire, A., & Lenormand, T. (2009) · Evolution · doi:10.1111/j.1558-5646.2009.00783.x · Cited by 32
Automatic relevance determination based hierarchical Bayesian MEG inversion in practice
Nummenmaa, A. R., Auranen, T., Hämäläinen, M., Jääskeläinen, I. P., Sams, M. E., Vehtari, A., & Lampinen, J. (2007) · NeuroImage · doi:10.1016/j.neuroimage.2007.04.021 · Cited by 37
“(16) (see Nummenmaa et al., in press; Gelman, 2006; Gelman et al., 2003, pp. 136, 390).”
Hierarchical Bayesian spatiotemporal analysis of revascularization odds using smoothing splines
Silva, G. L., Dean, C. B., Niyonsenga, T., & Vanasse, A. (2007) · Statistics in Medicine · doi:10.1002/sim.3094 · Cited by 41
“Gelman (2006) pointed 15 out that inference can become very sensitive for datasets in which low values of σ2 are possible, suggesting a half-t distribution for the variance component instead of an inverse gamma prior in hierarchical models.”
BackgroundThe network meta-analytic-predictive approach to non-inferiority trials
Schmidli, H., Wandel, S., & Neuenschwander, B. E. (2012) · Statistical Methods in Medical Research · doi:10.1177/0962280211432512 · Cited by 33
Almuqbel, M. M., Melzer, T. R., Myall, D. J., MacAskill, M. R., Pitcher, T. L., Livingston, L., Horne, K., Keenan, R. J., Dalrymple‐Alford, J. C., & Anderson, T. (2015) · Parkinsonism & Related Disorders · doi:10.1016/j.parkreldis.2015.11.008 · Cited by 34
“5) were used for subject-level and measurementlevel predictors and half-Cauchy priors (mean 0, scale 1) were used for variance parameters [29,30].”
Repeatable and heritable behavioural variation in a wild cooperative breeder
Edwards, H. A., Burke, T. A., & Dugdale, H. L. (2017) · Behavioral Ecology · doi:10.1093/beheco/arx013 · Cited by 31
Walrus haul-out and in water activity levels relative to sea ice availability in the Chukchi Sea
Jay, C. V., Taylor, R. L., Fischbach, A. S., et al. (2016) · Journal of Mammalogy · doi:10.1093/jmammal/gyw195 · Cited by 32
Bayesian Estimation of Random Coefficient Dynamic Factor Models
Song, H., & Ferrer, E. E. (2012) · Multivariate Behavioral Research · doi:10.1080/00273171.2012.640593 · Cited by 33
“To solve this potential problem, Gelman (2006) suggested using the half-t family when the number of cross-sectional units is small and in other settings where a weakly informative prior is desired.”
BackgroundMethodInfluentialRotational complexity increases cropping system output under poorer growing conditions
Bybee‐Finley, K. A., Muller, K. E., White, K. E., Cavigelli, M. A., Han, E., Schomberg, H. H., Snapp, S. S., Viens, F. G., Correndo, A. A., Deiss, L., et al. (2024) · One Earth · doi:10.1016/j.oneear.2024.07.008 · Cited by 41
“Hyperparameters for group-level standard deviation ( s a site ; s a mgmt ; s a replicate ; s c ; s d ) and the parameter for population-level standard deviation ( s in Equation 3) were specified as half non-standard student t priors, as recommended by Gelman 72 (Note S6).”
MethodBronner, A., Hénaux, V., Vergne, T., Vinard, J.-L., Morignat, E., Hendrikx, P., Calavas, D., & Gay, É. (2013) · PLoS ONE · doi:10.1371/journal.pone.0063246 · Cited by 44
The use of continuous data versus binary data in MTC models: A case study in rheumatoid arthritis
Schmitz, S., Adams, R., & Walsh, C. (2012) · BMC Medical Research Methodology · doi:10.1186/1471-2288-12-167 · Cited by 41
“A uniform distribution was chosen for the between trial standard deviation parameter σ δ/α , as proposed by Gelman [29].”
Chargé, R., Sorci, G., Jalme, M. S., Lesobre, L., Hingrat, Y., Lacroix, F., & Teplitsky, C. (2014) · Evolutionary Applications · doi:10.1111/eva.12150 · Cited by 41
“We assessed two priors for variances ( V A , V PE and V YEAR ) for each analysis: (i) a parameter-expanded prior (Gelman 2006), which is weakly informative prior of the shape ( V = 1, g = 1, a . l = 0 and a .”
Bally, M., Dendukuri, N., Sinclair, A., Ahern, S. P., Poisson, M., & Brophy, J. M. (2012) · International Journal of Antimicrobial Agents · doi:10.1016/j.ijantimicag.2012.08.004 · Cited by 29
Bayesball: A Bayesian hierarchical model for evaluating fielding in major league baseball
Jensen, S. T., Shirley, K. E., & Wyner, A. J. (2009) · The Annals of Applied Statistics · doi:10.1214/08-aoas228 · Cited by 36
Space–time latent component modeling of geo‐referenced health data
Lawson, A. B., Song, H., Cai, B., et al. (2010) · Statistics in Medicine · doi:10.1002/sim.3917 · Cited by 31
Use of Wishart Prior and Simple Extensions for Sparse Precision Matrix Estimation
Kuismin, M., & Sillanpää, M. J. (2016) · PLoS ONE · doi:10.1371/journal.pone.0148171 · Cited by 32
Rankin, R. W., Nicholson, K. E., Allen, S. J., et al. (2016) · Frontiers in Marine Science · doi:10.3389/fmars.2016.00025 · Cited by 31
“This follows the work of Gelman (2006) who popularized the scaled half Studentt distribution for the dispersion parameters in Hierarchical Bayesian models (but not necessarily for CMR).”
Hybrid larch heterosis: for which traits and under which genetic control?
Marchal, A., Muñoz, F., Millier, F., Sánchez, L., & Pâques, L. E. (2017) · Tree Genetics & Genomes · doi:10.1007/s11295-017-1177-1 · Cited by 30
Ramakers, J. J. C., Čulina, A., Visser, M. E., & Gienapp, P. (2018) · Nature Ecology & Evolution · doi:10.1038/s41559-018-0577-4 · Cited by 35
“This is in agreement with previously voiced concerns that the IW prior may behave poorly when true variance is close to zero66, 69, 70.”
Baek, E., Beretvas, S. N., Noortgate, W. V. D., & Ferron, J. M. (2019) · The Journal of Experimental Education · doi:10.1080/00220973.2018.1527280 · Cited by 30
“For example, Gelman (2006) demonstrated the impact of various proposed noninformative prior distributions for the Level 2 variance parameters in multilevel models.”
BackgroundMethodInfluentialA Bayesian Monte Carlo approach for predicting the spread of infectious diseases
Stojanović, O., Leugering, J., Pipa, G., Ghozzi, S., & Ullrich, A. (2019) · PLoS ONE · doi:10.1371/journal.pone.0225838 · Cited by 31
“A half-Cauchy distribution is used as a weakly informative prior [15] to enforce positivity of the dispersion parameter of the residual Negative Binomial distribution.”
MethodBorlongan, I. A., Arita, R., Nishihara, G. N., & Terada, R. (2020) · Journal of Applied Phycology · doi:10.1007/s10811-020-02266-2 · Cited by 32
“A halfcauchy prior was also placed on the scale parameter of the models (Gelman 2004, 2006).”
MethodHabitat Features Predict Carrying Capacity of a Recovering Marine Carnivore
Tinker, M. T., Yee, J. L., Laidre, K. L., Hatfield, B. B., Harris, M. D., Tomoleoni, J. A., Bell, T. W., Saarman, E. T., Carswell, L. P., & Miles, A. K. (2021) · Journal of Wildlife Management · doi:10.1002/jwmg.21985 · Cited by 38
“…based on biological feasibility but having no information specific to the analysis), including Cauchy priors for fixed‐ effect parameters in equations 4 and 5, half‐Cauchy priors for all variance and dispersion parameters (Gelman 2006, Gelman et al. 2008), and gamma priors for D* and θ (Table 1).”
MethodWilliams, D. R., Martin, S. R., Liu, S., & Rast, P. (2020) · European Journal of Psychological Assessment · doi:10.1027/1015-5759/a000624 · Cited by 42
Rethinking the Funding Line at the Swiss National Science Foundation: Bayesian Ranking and Lottery
Heyard, R., Ott, M., Salanti, G., & Egger, M. (2022) · Statistics and Public Policy · doi:10.1080/2330443x.2022.2086190 · Cited by 41
“Especially for the variance parameters, alternative prior distributions could be investigated, like half-normal and halfCauchy priors, whereas inverse-Gamma priors are generally not recommended [Gelman, 2006].”
Methodbamlss: A Lego Toolbox for Flexible Bayesian Regression (and Beyond)
Umlauf, N., Klein, N., Simon, T., & Zeileis, A. (2021) · Journal of Statistical Software · doi:10.18637/jss.v100.i04 · Cited by 56
Çınar, İ., & Köklü, M. (2021) · Selcuk Journal of Agricultural and Food Sciences · doi:10.15316/sjafs.2021.252 · Cited by 31
Meyer, S., Robertson, B. C., Chilvers, B. L., & Krkošek, M. (2015) · Marine Biology · doi:10.1007/s00227-015-2695-8 · Cited by 39
“Furthermore, uniform prior distributions were provided for temporal standard deviations of survival on the logit scale σlogit(φk,t) (Gelman 2006).”
MethodPopulation trends of female Elephant Seals breeding on the Courbet Peninsula, îles Kerguelen
Authier, M., Delord, K., & Guinet, C. (2010) · Polar Biology · doi:10.1007/s00300-010-0881-1 · Cited by 35
“Weakly informative priors (Half-Cauchy and Cauchy priors for variance and other parameters, respectively) were used (Gelman 2006; Gelman et al. 2008; Fùquene et al. 2009).”
MethodHierarchical Dirichlet Processes with Random Effects
Kim, S., & Smyth, P. (2007) · The MIT Press eBooks · doi:10.7551/mitpress/7503.003.0092 · Cited by 31
“We define prior distributions forΨbm andψ 2 hm as a half normal distribution with a0 mean and a variance as suggested by Gelman (2006).”
MethodBayesian approaches to the quantitative genetic analysis of natural populations
Morrissey, M. B., de Villemereuil, P., Doligez, B., & Giménez, O. (2014) · Oxford University Press eBooks · doi:10.1093/acprof:oso/9780199674237.003.0014 · Cited by 37
A nonparametric Bayesian methodology for regression discontinuity designs
Branson, Z., Rischard, M., Bornn, L., & Miratrix, L. W. (2019) · Journal of Statistical Planning and Inference · doi:10.1016/j.jspi.2019.01.003 · Cited by 42
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.