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 1,701–1,800 of 4,089
Abundant artificial grasslands around forests increase the deer impact on forest vegetation
Takarabe, K., & Iijima, H. (2020) · European Journal of Forest Research · doi:10.1007/s10342-020-01262-y · Cited by 13
“Prior distribution of CO is a vague uniform distribution, ranging from 0 to 100 (Gelman 2006).”
Lewis‐Beck, C., Walker, V. A., Niemi, J., Caragea, P. C., & Hornbuckle, B. K. (2020) · Remote Sensing · doi:10.3390/rs12050827 · Cited by 14
“For the standard deviations of the two covariance matrices (Λμ, Λθ), as well as the standard deviations seasonal stochastic error terms (νs), we specify independent half-Cauchy distributions [36,37].”
MethodBiswas, N., Bhattacharya, A., Jacob, P. E., & Johndrow, J. E. (2022) · Journal of the Royal Statistical Society Series B (Statistical Methodology) · doi:10.1111/rssb.12495 · Cited by 12
“On the statistical estimation side, Half-t ( 𝜈 ) priors have been long proposed (Carvalho et al., 2009, 2010; Gelman, 2006) for Bayesian hierarchical models. van der Pas et al. (2014, 2017) have established (near) minimax optimal estimation for the Horseshoe ( 𝜈 = 1) in the normal means model.”
BackgroundMethodSimplifying Small Area Estimation With rFIA: A Demonstration of Tools and Techniques
Stanke, H., Finley, A. O., & Domke, G. M. (2022) · Frontiers in Forests and Global Change · doi:10.3389/ffgc.2022.745874 · Cited by 12
“We choose weakly informative normal priors for α (i.e., mean 50, standard deviation 250) and β (i.e., mean 0, standard deviation 100), and weakly informative half student-t priors for all variance terms (i.e., mean 0, scale 100, 3 degrees of freedom) (Gelman, 2006).”
BackgroundMethodBroglio, K. R., Zhang, F., Yu, B., Marshall, J., Wang, F., Bennett, M., & Viele, K. (2021) · Statistics in Biopharmaceutical Research · doi:10.1080/19466315.2021.2008484 · Cited by 14
A randomness perspective on intelligence processes
Kang, I., de Boeck, P. A. L., & Partchev, I. (2022) · Intelligence · doi:10.1016/j.intell.2022.101632 · Cited by 12
A Bayesian Approach to Estimating Reciprocal Effects with the Bivariate STARTS Model
Lüdtke, O., Robitzsch, A., & Ulitzsch, E. (2022) · Multivariate Behavioral Research · doi:10.1080/00273171.2022.2039585 · Cited by 19
Quantifying the impacts of 166 years of land cover change on lowland bird communities
Kitazawa, M., Yamaura, Y., Senzaki, M., Hanioka, M., Ohashi, H., Oguro, M., Matsui, T., & Nakamura, F. (2022) · Proceedings of the Royal Society B Biological Sciences · doi:10.1098/rspb.2022.0338 · Cited by 13
“Weuseduninformativenormaldistributions (0, 100(2)) for the mean value of bi1 8 and half-Cauchy distributions (0, 5) for any variance of coefficients as weakly informative prior distributions [49].”
BackgroundMultilevel time-series models for small area estimation at different frequencies and domain levels
Boonstra, H. J., & van den Brakel, J. (2022) · The Annals of Applied Statistics · doi:10.1214/21-aoas1592 · Cited by 11
Chinembiri, T. S., Mutanga, O., & Dube, T. J. (2022) · Ecological Informatics · doi:10.1016/j.ecoinf.2022.101934 · Cited by 15
Giordano, K. R., Saber, M., Green, T. R. F., Rojas-Valencia, L. M., Ortiz, J. B., Murphy, S. M., Lifshitz, J., & Rowe, R. K. (2023) · Neurotrauma Reports · doi:10.1089/neur.2022.0092 · Cited by 15
Kim, J., Lawson, A. B., Neelon, B., Korte, J. E., Eberth, J. M., & Chowell, G. (2023) · BMC Medical Research Methodology · doi:10.1186/s12874-023-01987-5 · Cited by 15
“For variance terms, we further assume weakly-informative prior distributions, and standard deviations follow uniform prior distributions recommended by Gelman [26].”
MethodThe race between education and technology in Chile and its impact on the skill premium
Campos-González, J., & Balcombe, K. (2023) · Economic Modelling · doi:10.1016/j.econmod.2023.106616 · Cited by 14
Bayesian Hierarchical/Multilevel Models for Inference and Prediction Using Cross-System Lake Data
Stow, C. A., Lamon, E. C., Qian, S. S., et al. (2008) · Springer eBooks · doi:10.1007/978-0-387-77942-3_5 · Cited by 12
“In such instances, Gelman (2006) recommends using a uniform prior for instead of a gamma prior for 1/ 2 to ensure that the prior is truly noninformative.”
BackgroundMethodChelgren, N. D., Samora, B., Adams, M. J., et al. (2011) · Herpetological conservation and biology · Cited by 12
Kazembe, L. N. (2013) · Spatial and Spatio-temporal Epidemiology · doi:10.1016/j.sste.2013.07.002 · Cited by 14
Augmented mixed models for clustered proportion data
Bandyopadhyay, D., Soto, D. M. G., & Lachos, V. H. (2014) · Statistical Methods in Medical Research · doi:10.1177/0962280214561093 · Cited by 11
Bayesian Semiparametric Intensity Estimation for Inhomogeneous Spatial Point Processes
Yu, Y., & Loh, J. M. (2010) · Biometrics · doi:10.1111/j.1541-0420.2010.01531.x · Cited by 12
“Other priors, such as the Pareto prior (Yue and Speckman, 2010) and half-Cauchy prior (Gelman, 2006), are also good options.”
BackgroundSupervised Bayesian latent class models for high‐dimensional data
DeSantis, S. M., Houseman, E. A., Coull, B. A., et al. (2012) · Statistics in Medicine · doi:10.1002/sim.4448 · Cited by 12
Choi, I.-H., & Wilson, M. R. (2014) · Educational and Psychological Measurement · doi:10.1177/0013164414522124 · Cited by 12
“However, given that the specification of the prior distributions could have substantial impacts on estimation (Gelman, 2006), it is worth investigating more deeply the use of different prior distributions.”
BackgroundGeneralized Dynamic Factor Models for Mixed-Measurement Time Series
Cui, K., & Dunson, D. B. (2012) · Journal of Computational and Graphical Statistics · doi:10.1080/10618600.2012.729986 · Cited by 13
“PX has been introduced as a useful approach for both introducing new families of prior distributions and improving computational efficiency (Liu and Wu 1999; Gelman 2004; Kinney and Dunson 2007).”
MethodSraj, I., Zedler, S. E., Knio, O., et al. (2016) · Monthly Weather Review · doi:10.1175/mwr-d-15-0394.1 · Cited by 12
“(10) Unauthenticated | Downloaded 06/15/22 08:01 PM UTC The prior p(S) follows a Gamma distribution (Wang and Zabaras 2005; Gelman et al. 2004; Gelman 2006; JH) that depends on scale and rate parameters a and b as follows: p(S)5 8>< >: ba G(a) Sa21 exp(2bS) , S. 0 0, otherwise.”
BackgroundCasals, M., Langohr, K., Carrasco, J. L., & Rönnegård, L. (2015) · Sort-statistics and Operations Research Transactions · Cited by 13
Ruggeri, F., Sawlan, Z., Scavino, M., & Tempone, R. F. (2016) · Bayesian Analysis · doi:10.1214/16-ba1007 · Cited by 12
“The half-Cauchy prior for η was chosen because it is a practical prior for scale parameters in hierarchical models (Polson and Scott, 2012) (Gelman, 2006).”
MethodPham, L., Carvalho, L. E., Schaus, S. E., & Kolaczyk, E. D. (2015) · Journal of the American Statistical Association · doi:10.1080/01621459.2015.1110523 · Cited by 21
“(7) Since an inverse gamma distribution with shape α corresponds to a χ2 distribution with 2α degrees of freedom (Gelman 2006), the prior provides 2ζ pseudo-observations and the conditional in (7) has 2ζ + n observations.”
BackgroundA Bayesian hierarchical surrogate outcome model for multiple sclerosis
Pozzi, L., Schmidli, H., & Ohlssen, D. I. (2016) · Pharmaceutical Statistics · doi:10.1002/pst.1749 · Cited by 13
Spatio-temporal relative survival of breast and colorectal cancer in Queensland, Australia 2001–2011
Cramb, S. M., Mengersen, K. L., & Baade, P. D. (2016) · Spatial and Spatio-temporal Epidemiology · doi:10.1016/j.sste.2016.08.002 · Cited by 12
Flexible shrinkage estimation of subgroup effects through Dirichlet process priors
Gamalo, M., Tiwari, R. C., & LaVange, L. M. (2016) · Journal of Biopharmaceutical Statistics · doi:10.1080/10543406.2016.1226327 · Cited by 13
Item Cloning Variation and the Impact on the Parameters of Response Models
Lathrop, Q. N., & Cheng, Y. (2016) · Psychometrika · doi:10.1007/s11336-016-9513-1 · Cited by 16
“A Newcomer” versus “First Mover”: Retail Location Strategy for Differentiation
Lee, J., & Kim, Y. (2017) · The Professional Geographer · doi:10.1080/00330124.2017.1310621 · Cited by 12
“As Gelman (2006) suggested, an exact positive half normal distribution NC 1 .0; 10/ is set to the standard deviation of spatial random effect (Bernardinelli et al. 1995; Y. Kim 2007; Lawson 2013; Law, Quick, and Chan 2014; G. Li et al. 2014; L. Li 2014; Yin et al. 2014).”
BackgroundSampling Strategies for Fast Updating of Gaussian Markov Random Fields
Brown, D. A., McMahan, C. S., & Self, S. W. (2019) · The American Statistician · doi:10.1080/00031305.2019.1595144 · Cited by 18
“It has been observed that an inverse gamma prior on τ 2 sometimes can yield undesirable behavior in the posterior (Gelman, 2006); but our focus is on sampling the random field and thus we use this prior simply for convenience.”
The shapes and alignments of the satellites of the Milky Way and Andromeda
Sanders, J. L., & Evans, N. W. (2017) · Monthly Notices of the Royal Astronomical Society · doi:10.1093/mnras/stx2116 · Cited by 12
“Instead, we heed the advice of Gelman (2006) and adopt a half-Cauchy distribution as a weakly informative prior.”
MethodMa, Q., Xiong, L., Li, Y., Li, S., & Xu, C. (2017) · The Science of The Total Environment · doi:10.1016/j.scitotenv.2017.09.235 · Cited by 14
“Therefore, we derived prior from the halfCauchy distribution, whose appealing feature is that it only provides a weak prior and encapsulates some basic constraints on parameters, instead of concentratingmore of prior close to zero (Gelman, 2006).”
MethodApproximate Bayesian evaluations of measurement uncertainty
Possolo, A. M., & Bodnar, O. (2018) · Metrologia · doi:10.1088/1681-7575/aaa5be · Cited by 12
“And we will model τ and σ as values of random variables with half-Cauchy distributions as recommended by Gelman [14], the former with median equal to the standard deviation of the bottle averages, the latter with median equal to the median of the within-bottle standard deviations.”
MethodEffects of Temperature and Spatial Scale on Rio Grande Cutthroat Trout Growth and Abundance
Huntsman, B. M., Martin, R. W., & Patten, K. A. (2018) · Transactions of the American Fisheries Society · doi:10.1002/tafs.10051 · Cited by 12
Bayesian semiparametric modeling for HIV longitudinal data with censoring and skewness
Castro, L. M., Wang, W., Lachos, V. H., et al. (2018) · Statistical Methods in Medical Research · doi:10.1177/0962280218760360 · Cited by 12
Risser, M. D., Calder, C. A., Berrocal, V. J., & Berrett, C. (2019) · The Annals of Applied Statistics · doi:10.1214/18-aoas1204 · Cited by 11
“Uniform priors for the variance parameters τ 2 and σ 2 are used in place of more traditional conjugate inverse-Gamma priors to avoid prior bias in the case where an individual segment contains a small number of observations (following Gelman (2006)).”
BackgroundMethodInfluentialHepler, S., McKnight, E., Bonny, A., & Kline, D. (2019) · Epidemiology · doi:10.1097/ede.0000000000000978 · Cited by 12
“The prior distributions specified by (8) imply π(τ (2)) ∝ τ−1, π(τ 2 D) ∝ τ−1 D , π(σ(2) D) ∝ σ −1 D and π(σ 2 T ) ∝ σ−1 T (25).”
Rotejanaprasert, C., Lawson, A. T. D.-D., Rossow, H., et al. (2018) · BMC Medical Research Methodology · doi:10.1186/s12874-018-0532-8 · Cited by 12
Hund, L., Schroeder, B., Rumsey, K. N., & Huerta, G. (2018) · Reliability Engineering & System Safety · doi:10.1016/j.ress.2018.07.017 · Cited by 13
Lee, J., Son, J., Zhou, S., & Chen, Y. (2020) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2019.2959605 · Cited by 12
“The IG specifications are adopted from the literature [16], [28].”
MethodHierarchical Bayesian CMB component separation with the No-U-Turn Sampler
Grumitt, R. D. P., Jew, L. R. P., & Dickinson, C. (2020) · Monthly Notices of the Royal Astronomical Society · doi:10.1093/mnras/staa1857 · Cited by 16
“In doing so, we are able to provide a faithful generative description of the underlying foreground emission, whilst reducing the propensity for fitting noisy outliers when assuming total independence between pixel spectral parameters (Gelman 2006b; Gelman & Hill 2007).”
BackgroundInfluentialUncovering drivers of dose-dependence and individual variation in malaria infection outcomes
Kamiya, T., Greischar, M. A., Schneider, D. S., & Mideo, N. (2020) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1008211 · Cited by 12
“We assumed that the individual deviations, u k , i and v k , i , are samples of a normal distribution with standard deviations, σ u , k and σ v , k that we estimated from data [74].”
Pérez‐Panadés, J., Botella-Rocamora, P., & Martínez‐Beneito, M. A. (2020) · International Journal of Health Geographics · doi:10.1186/s12942-020-00251-z · Cited by 12
Nations, J. A., Mount, G. G., Morere, S. M., Achmadi, A. S., Rowe, K. C., & Esselstyn, J. A. (2020) · Evolution · doi:10.1111/evo.14156 · Cited by 12
Rehman, H., Chandra, N., Hosseini‐Baharanchi, F. S., et al. (2021) · Journal of Applied Statistics · doi:10.1080/02664763.2021.1882407 · Cited by 12
Bayesian Inference under Small Sample Sizes Using General Noninformative Priors
He, J., Wang, W., Huang, M., Wang, S., & Guan, X. (2021) · Mathematics · doi:10.3390/math9212810 · Cited by 16
“It was further argued that the resulting inference based on it for cases where σ2 is estimated near zero was the case for which classical and Bayesian inferences differed the most [35].”
BackgroundInfluentialOn Bayesian modeling of censored data in JAGS
Qi, X., Zhou, S., & Plummer, M. (2022) · BMC Bioinformatics · doi:10.1186/s12859-021-04496-8 · Cited by 22
“Durvalumab) effect on the incidence of AE, we begin with modeling drug effects without any link function as Model C, and then extend to specify half-Cauchy prior [46] to the standard deviation of drug effect with logit, cloglog, and probit link functions in Model D-F, respectively.”
MethodHigh habitat invasibility unveils the invasiveness potential of water frogs
Denoël, M., Duret, C., Lorrain‐Soligon, L., Padilla, P., Pavis, J., Pille, F., Tendron, P., Ficetola, G. F., & Falaschi, M. (2022) · Biological Invasions · doi:10.1007/s10530-022-02849-9 · Cited by 14
“The tau parameter followed a half-Cauchy distribution (Gelman 2006); the prior for the precision parameter of the ICAR model was taken from a gamma distribution with mean = 0 and precision = 0.”
MethodInfluentialRöver, C., Sturtz, S., Lilienthal, J., Bender, R., & Friede, T. (2023) · Statistics in Medicine · doi:10.1002/sim.9731 · Cited by 17
Fossil leaves reveal drivers of herbivore functional diversity during the Cenozoic
Albrecht, J., Wappler, T., Fritz, S. A., & Schleuning, M. (2023) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2300514120 · Cited by 12
Robinson, T. J., Anderson‐Cook, C. M., & Hamada, M. S. (2009) · Technometrics · doi:10.1198/tech.2009.0006 · Cited by 12
“For σδ , we use the diffuse prior distribution Uniform(0, 100) for fσδ (·), as suggested by Gelman (2006) for the standard deviation of the random-effects distribution in hierarchical models.”
MethodA Bayesian regression model for multivariate functional data
Rosen, O., & Thompson, W. K. (2009) · Computational Statistics & Data Analysis · doi:10.1016/j.csda.2009.03.026 · Cited by 12
Chen, M., Das, S., Kim, S., & Warren, N. D. (2008) · Bayesian Analysis · doi:10.1214/08-ba308 · Cited by 13
“We note that other priors for σ2k, σ 2 τ , and σ 2 τ∗ like those discussed by Gelman (2006) can also be considered.”
BackgroundParameter Expansion and Efficient Inference
Lewandowski, A., Liu, C., & Wiel, S. V. (2010) · Statistical Science · doi:10.1214/10-sts348 · Cited by 14
Carcinogen metabolism, cigarette smoking, and breast cancer risk: a Bayes model averaging approach
Stephenson, N., Beckmann, L., & Chang, J. C. (2010) · Epidemiologic Perspectives & Innovations · doi:10.1186/1742-5573-7-10 · Cited by 12
Varying-coefficient models for longitudinal processes with continuous-time informative dropout
Su, L., & Hogan, J. W. (2009) · Biostatistics · doi:10.1093/biostatistics/kxp040 · Cited by 12
Estimating Riverwide Abundance of Juvenile Fish Populations: How Much Sampling is Enough?
Korman, J., Schick, J., & Mossop, B. (2016) · North American Journal of Fisheries Management · doi:10.1080/02755947.2015.1114542 · Cited by 12
Efficient simulation from a gamma distribution with small shape parameter
Liu, C., Martin, R., & Syring, N. (2016) · Computational Statistics · doi:10.1007/s00180-016-0692-0 · Cited by 12
A sparse hierarchical Bayesian model for detecting relevant antigenic sites in virus evolution
Davies, V., Reeve, R. E., Harvey, W. T., et al. (2017) · Computational Statistics · doi:10.1007/s00180-017-0730-6 · Cited by 12
“Finally the overparameterisation can improve sampling by reducing the dependence between parameters in the hierarchical model leading to improved MCMC convergence (Gelman 2004).”
BackgroundModelling Subject Variability in the Spatial and Temporal Characteristics of Functional ModesPreprint
Harrison, S. J., Bijsterbosch, J. D., Segerdahl, A. R., et al. (2019) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/544817 · Cited by 12
Nalborczyk, L., Batailler, C., Lœvenbruck, H., Vilain, A., & Bürkner, P. (2019) · PsyArXiv (OSF Preprints) · doi:10.31234/osf.io/guhsa · Cited by 13
Allogeneic Bone Marrow‐Derived Mesenchymal Stem Cells for Parkinson's Disease: A Randomized Trial
Schiess, M. C., Suescun, J., Lemus, J. D. L., Green, C. E., Thomas, T. S., Shahnawaz, M., Tharp, E., Satani, N., Saltarrelli, J. G., Adams, C., et al. (2025) · Movement Disorders · doi:10.1002/mds.70028 · Cited by 16
The origins of regional integration: Untangling the effect of trade on judicial cooperation
Dyèvre, A., & Lampach, N. (2018) · International Review of Law and Economics · doi:10.1016/j.irle.2018.08.003 · Cited by 12
Variance partitioning in spatio-temporal disease mapping models
Franco‐Villoria, M., Ventrucci, M., & Rue, H. (2022) · Statistical Methods in Medical Research · doi:10.1177/09622802221099642 · Cited by 11
“By scaling the structure matrix of the temporal and spatial random effects, the structure matrix of the interaction, defined as a Kronecker IGMRF, is automatically scaled.”
BackgroundMethodInfluentialBayesian hierarchical response time modelling—A tutorial
Koenig, C., Becker, B., & Ulitzsch, E. (2023) · British Journal of Mathematical and Statistical Psychology · doi:10.1111/bmsp.12302 · Cited by 14
Wigle, A., Béliveau, A., Blackmore, D., et al. (2024) · ACS ES&T Air · doi:10.1021/acsestair.4c00030 · Cited by 11
Moffat, R., Caruana, N., & Cross, E. S. (2024) · Open Biology · doi:10.1098/rsob.230382 · Cited by 9
“We set weakly informed priors to impose a constrained distribution on our expected results, thereby acknowledging the limits of our knowledge as to our expected results, allowing for possible large effects and allowing the data to dominate the posterior distribution structure [89,90].”
MethodLocal and global trend Bayesian exponential smoothing models
Smyl, S., Bergmeir, C., Dokumentov, A., et al. (2024) · International Journal of Forecasting · doi:10.1016/j.ijforecast.2024.03.006 · Cited by 11
Bayesian Hierarchical Models for Subgroup Analysis
Wang, Y., Tu, W., Koh, W. J. H., Travis, J. W., Abugov, R., Hamilton, K., Zheng, M., Crackel, R., Bonangelino, P., & Rothmann, M. D. (2024) · Pharmaceutical Statistics · doi:10.1002/pst.2424 · Cited by 11
A random effects meta-analysis model with Box-Cox transformation
Yamaguchi, Y., Maruo, K., Partlett, C., & Riley, R. D. (2017) · BMC Medical Research Methodology · doi:10.1186/s12874-017-0376-7 · Cited by 11
Liang, X., Yang, Y., & Huang, J. (2017) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705511.2017.1410706 · Cited by 16
“Furthermore, hierarchical Bayesian methods have drawn attention recently in SEM regularization, by which hyper-priors (Gelman, 2006a) can be assigned to penalized parameters.”
BackgroundMethodBayesian analysis of randomized controlled trials
Bautista, J. R., Pavlakis, A., & Rajagopal, A. (2018) · International Journal of Eating Disorders · doi:10.1002/eat.22928 · Cited by 14
Extending Bayesian back-calculation to estimate age and time specific HIV incidence
Brizzi, F., Birrell, P., Plummer, M., et al. (2019) · Lifetime Data Analysis · doi:10.1007/s10985-019-09465-1 · Cited by 11
“To reflect a lack of prior knowledge and a weak preference towards smooth curves, diffuse half-t prior distributions with 2 degrees of freedom and scale parameter 200 are chosen so that 95% of the prior density lies in the [0, 400] region (Gelman 2006).”
BackgroundThrive at Work Wellbeing Programme Collaboration (2019) · BMC Public Health · doi:10.1186/s12889-019-6582-y · Cited by 11
Uto, M., Miyazawa, Y., Kato, Y., Nakajima, K., & Kuwata, H. (2020) · International Journal of Artificial Intelligence in Education · doi:10.1007/s40593-019-00189-9 · Cited by 15
“The inverse gamma distribution IG(g 1 / 2 , g 2 / 2 ) , a type of conjugate prior, is often used as the prior distribution of normal distribution variance σ 2 sf (Gelman 2006), where g 1 and g 2 are hyperparameters and are generally small positive real numbers, such as g 1 = g 2 = 0.01.”
MethodInfluentialJung, H., Senf, C., Beudert, B., & Krueger, T. (2021) · Water Resources Research · doi:10.1029/2020wr027264 · Cited by 13
“Weakly informative Normal (0,5) prior distributions were assigned to m. Weakly informative Inverse-gamma (0.01, 0.01) priors were assigned to m, and as these parameters are bound to be positive (Gelman, 2006).”
MethodInfluentialVanhatalo, J. P., Hartmann, M., & Veneranta, L. (2020) · Jukuri (Natural Resources Institute Finland (Luke)) · Cited by 11
Synthetic Microdata for Establishment Surveys Under Informative Sampling
Kim, H. J., Drechsler, J., & Thompson, K. J. (2020) · Journal of the Royal Statistical Society Series A (Statistics in Society) · doi:10.1111/rssa.12622 · Cited by 14
Sato, Y., Nagoe, H., Nishihara, G. N., & Terada, R. (2021) · Journal of Applied Phycology · doi:10.1007/s10811-021-02535-8 · Cited by 11
“…chains were sampled to provide at least 5000 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).”
MethodAssessment and novel application ofN‐mixture models for aerial surveys of wildlife
Christensen, S. A., Farr, M. T., & Williams, D. M. (2021) · Ecosphere · doi:10.1002/ecs2.3725 · Cited by 11
Campbell, H., & Gustafson, P. (2021) · Epidemiology and Infection · doi:10.1017/s0950268821002405 · Cited by 11
Mapping the perception-space of facial expressions in the era of face masks
Verroca, A., de Rienzo, C. M., Gambarota, F., & Sessa, P. (2022) · Frontiers in Psychology · doi:10.3389/fpsyg.2022.956832 · Cited by 12
“We decided to use weakly informative priors for regression parameters (Gelman, 2006; Gelman et al., 2017).”
MethodDefining Hydrogeological Site Similarity with Hierarchical Agglomerative Clustering
Kawa, N., Cucchi, K., Rubin, Y., Attinger, S., & Heße, F. (2022) · Ground Water · doi:10.1111/gwat.13261 · Cited by 4
Forecasting sub-national trends in COVID-19 vaccine uptake in the UK before vaccine rollout
de Figueiredo, A. (2022) · Scientific Reports · doi:10.1038/s41598-022-25354-4 · Cited by 23
Analysis of N‐of‐1 trials using Bayesian distributed lag model with autocorrelated errors
Liao, Z., Qian, M., Kronish, I. M., & Cheung, Y. K. K. (2023) · Statistics in Medicine · doi:10.1002/sim.9676 · Cited by 12
“σ −2E∗E∗ + σ−2 φ I )−1 σ −2E∗ε∗, ( σ −2E∗E∗ + σ−2 φ I )−1] 1Sφ (φ), (15)”
High-dimensional order-free multivariate spatial disease mapping
Fuenzalida, G. V., Adín, A., Goicoa, T., & Ugarte, M. D. (2023) · Statistics and Computing · doi:10.1007/s11222-023-10263-x · Cited by 15
Ward, E. J., Pess, G. R., Anlauf‐Dunn, K. J., & Jordan, C. E. (2012) · Canadian Journal of Fisheries and Aquatic Sciences · doi:10.1139/f2012-096 · Cited by 11
Prior information reduces uncertainty about the consequences of deer overabundance on forest birds
Martin, T. G., Arcese, P., Kuhnert, P., Gaston, A. J., & Martin, J. (2013) · Biological Conservation · doi:10.1016/j.biocon.2013.05.017 · Cited by 11
Rejoinder: sifting through model space
Heisey, D. M., Osnas, E. E., Cross, P. C., Joly, D. O., Langenberg, J. A., & Miller, M. W. (2010) · Ecology · doi:10.1890/10-0894.1 · Cited by 13
Kazembe, L. N. (2013) · PLoS ONE · doi:10.1371/journal.pone.0073500 · Cited by 10
Temporal dynamics of fine roots under long‐term exposure to elevated CO2 in the Mojave Desert
Sonderegger, D. L., Ogle, K., Evans, R. D., Ferguson, S. D., & Nowak, R. S. (2013) · New Phytologist · doi:10.1111/nph.12128 · Cited by 10
“; vT ;RÞ for each random effect were assigned folded Cauchy priors to avoid posterior distributions with unrealistically heavy right tails (Gelman, 2006).”
MethodLatent variable models with nonparametric interaction effects of latent variables
Song, X., Lu, Z., & Feng, X. (2013) · Statistics in Medicine · doi:10.1002/sim.6065 · Cited by 11
Bayesian inference of protein structure from chemical shift data
Bratholm, L. A., Christensen, A. S., Hamelryck, T., & Jensen, J. H. (2015) · PeerJ · doi:10.7717/peerj.861 · Cited by 11
“…maximum a posteriori estimator of σ2j does not equal zero in the limit of79 χ2j → 0 with a non-zero choice of β :80 σ2j,MAP = 2β +χ2j (X) 2α +2+n j (13) In all the simulations where σ j was sampled we use Eqn 12 and α = β = 0.001 (Gelman, 2006) unless81 stated otherwise.82 83 Marginal likelihood.”
InfluentialMeta‐regression with partial information on summary trial or patient characteristics
Hemming, K., Hutton, J. L., Maguire, M. G., & Marson, A. G. (2010) · Statistics in Medicine · doi:10.1002/sim.3848 · Cited by 15
Jiao, Y., O'Reilly, R., Smith, E. P., Orth, D. J., & Jardim, H. E. E. (2016) · ICES Journal of Marine Science · doi:10.1093/icesjms/fsw036 · Cited by 13
“Usually two types of prior distributions are used in a Bayesian analysis: non-informative and informative (Berger, 1985; Gelman et al., 2004; Gelman, 2006).”
MethodCommensurate Priors on a Finite Mixture Model for Incorporating Repository Data in Clinical Trials
Gajewski, B. J., Reese, C. S., Colombo, J. A., & Carlson, S. E. (2016) · Statistics in Biopharmaceutical Research · doi:10.1080/19466315.2015.1133453 · Cited by 11
“…is limited since it is comparable to trying to estimate a standard deviation with just two observations (Spiegelhalter, Abrams, and Myles 2004; Gelman 2006), The prior selected for τk, the commensurability parameter, is a gamma distribution (Model 7); however, in Hobbs, Sargent, and Carlin…”
BackgroundFlexible Bayesian P-splines for smoothing age-specific spatio-temporal mortality patterns
Goicoa, T., Adín, A., Etxeberria, J., et al. (2017) · Statistical Methods in Medical Research · doi:10.1177/0962280217726802 · Cited by 11
Kaneko, Y., Miyazaki, S., & Yada, K. (2017) · Procedia Computer Science · doi:10.1016/j.procs.2017.08.225 · Cited by 11
Ranlund, Å., & Victorsson, J. (2017) · Forest Ecology and Management · doi:10.1016/j.foreco.2017.10.030 · Cited by 14
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.