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 701–800 of 4,089
Replication and Meta‐analysis of Genome‐Wide Association Studies
Dudbridge, F., & Newcombe, P. J. (2019) · Wiley eBooks · doi:10.1002/9781119487845.ch22 · Cited by 40
Ren, L., Kutaka, T. S., Chernyavskiy, P., Fan, J., & Li, X. (2020) · Contemporary Educational Psychology · doi:10.1016/j.cedpsych.2020.101845 · Cited by 41
Jara, A., Hanson, T., & Lesaffre, E. M. E. H. (2009) · Journal of Computational and Graphical Statistics · doi:10.1198/jcgs.2009.07062 · Cited by 40
“Other choices of prior distributions for the dispersion parameter τ were discussed in Gelman (2006) and can be easily implemented in our context.”
BackgroundMedeiros, F. A., Weinreb, R. N., Moore, G. H., et al. (2012) · Ophthalmology · doi:10.1016/j.ophtha.2011.10.003 · Cited by 40
Estimating the Relative Excess Risk Due to Interaction: A Bayesian Approach
Chu, H., Nie, L., & Cole, S. R. (2011) · Epidemiology · doi:10.1097/ede.0b013e318208750e · Cited by 40
Bell, A., Smith, J., Sabel, C. E., & Jones, K. (2016) · Journal of Quantitative Analysis in Sports · doi:10.1515/jqas-2015-0050 · Cited by 40
“The models were also estimated using Uniform priors for the variances, as suggested by Gelman (2006), but the results were almost identical and the substantive conclusions did not change.”
ResultBayes Factor Testing of Multiple Intraclass Correlations
Mulder, J., & Fox, J. (2018) · Bayesian Analysis · doi:10.1214/18-ba1115 · Cited by 35
“Furthermore, the (matrix) F prior is useful when estimating variance components (Gelman, 2006; Pérez et al., 2017) and for testing variances (Mulder and Pericchi, 2018).”
BackgroundMethodInfluentialWolak, M. E., Arcese, P., Keller, L. F., Nietlisbach, P., & Reid, J. M. (2018) · Evolution · doi:10.1111/evo.13575 · Cited by 42
Gaiser, R. A., Pessia, A., Ateeb, Z., Davanian, H., Moro, C. F., Alkharaan, H., Healy, K., Ghazi, S. H., Arnelo, U., Valente, R., et al. (2019) · Scientific Reports · doi:10.1038/s41598-019-46634-6 · Cited by 40
Establishing Meaningful Limits of Detection for Ion-Selective Electrodes and Other Nonlinear Sensors
Dillingham, P. W., Alsaedi, B. S. O., Granados-Fócil, S., et al. (2019) · ACS Sensors · doi:10.1021/acssensors.9b02133 · Cited by 40
Vicente, G., Goicoa, T., Rasines, P. F., & Ugarte, M. D. (2019) · Journal of the Royal Statistical Society Series A (Statistics in Society) · doi:10.1111/rssa.12545 · Cited by 52
“This agrees with the recommendation from Gelman (2006) about using non-informative uniform priors on the standard deviations.”
MethodZheng, S., Zhu, Y., Li, D. Q., Cao, Z.-J., Deng, Q.-X., & Phoon, K. (2020) · Geoscience Frontiers · doi:10.1016/j.gsf.2020.03.017 · Cited by 41
Bayesian inference and model comparison for asymmetric smooth transition heteroskedastic models
Gerlach, R., & Chen, C. W. S. (2008) · Statistics and Computing · doi:10.1007/s11222-008-9063-1 · Cited by 47
King, R., Bird, S. M., Overstall, A. M., Hay, G., & Hutchinson, S. J. (2013) · Journal of the Royal Statistical Society Series A (Statistics in Society) · doi:10.1111/rssa.12011 · Cited by 47
Lin, C.-Y., Hsu, C.-Y., Gunnell, D., Chen, Y.-Y., & Chang, S.-S. (2019) · Social Science & Medicine · doi:10.1016/j.socscimed.2018.12.011 · Cited by 39
“We checked the convergence of models by visual inspection of three chains and through examination of the Gelman-Rubin diagnostic (Gelman, 2006) (see Appendix figure 1).”
InfluentialData integration in causal inference
Shi, X., Pan, Z., & Miao, W. (2022) · Wiley Interdisciplinary Reviews Computational Statistics · doi:10.1002/wics.1581 · Cited by 77
“Bayesian framework can naturally facilitate the borrowing of prior information across data sources (Ibrahim and Chen, 2000; Gelman, 2006; Hobbs et al., 2011; Kaizer et al., 2018).”
Pick, J. L., Kasper, C., Allegue, H., Dingemanse, N. J., Dochtermann, N. A., Laskowski, K. L., Lima, M. R., Schielzeth, H., Westneat, D. F., WRIGHT, J. M., et al. (2023) · Methods in Ecology and Evolution · doi:10.1111/2041-210x.14200 · Cited by 45
Paciorek, C. J., Goring, S. J., Thurman, A. L., Cogbill, C. V., Williams, J. W., Mladenoff, D. J., Peters, J. A., Zhu, J., & McLachlan, J. S. (2016) · PLOS ONE · doi:10.1371/journal.pone.0150087 · Cited by 32
“Following [24] we use a uniform distribution on each σp parameter, with upper bound of 1000.”
MethodPropriety of posteriors in structured additive regression models: Theory and empirical evidence
Fahrmeir, L., & Kneib, T. (2008) · Journal of Statistical Planning and Inference · doi:10.1016/j.jspi.2008.05.036 · Cited by 50
“This prior corresponds to a special case of (aa), and is recommended as a standard choice in practical work by Gelman (2006).”
BackgroundPerception-based Personalization of Hearing Aids using Gaussian Processes and Active Learning
Nielsen, J., Nielsen, J., & Larsen, J. O. (2014) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2014.2377581 · Cited by 39
A Bayesian Approach to the Analysis of Gauge R&R Data
Weaver, B., Hamada, M. S., Vardeman, S. B., & Wilson, A. G. (2012) · Quality Engineering · doi:10.1080/08982112.2012.702381 · Cited by 41
“Because our computations are performed in WinBUGS, we approximate the Uniform(0, 1) distribution with the Uniform(0, A) distribution for A large as recommended by Gelman (2006).”
BackgroundMethodRota, C. T., Rumble, M. A., Millspaugh, J. J., Lehman, C. P., & Kesler, D. C. (2013) · Forest Ecology and Management · doi:10.1016/j.foreco.2013.10.048 · Cited by 46
Hadfield, J. D., Heap, E. A., Bayer, F., Mittell, E. A., & Crouch, N. M. A. (2013) · Evolution · doi:10.1111/evo.12143 · Cited by 42
“…expanded priors were used for the remaining variance components to give either scaled F-distributions with numerator and denominator degrees of freedom set to one and a scale parameter of 1000 (Gelman 2006), or in the case of the ordinal models a χ21 distribution (de Villemereuil et al. 2012).”
MethodJeon, B.-S., Ruszczycky, M. W., Russell, W. K., et al. (2017) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1710496114 · Cited by 39
Individual Differences in Motor Noise and Adaptation Rate Are Optimally Related
van der Vliet, R., Frens, M. A., de Vreede, L., Jonker, Z. D., Ribbers, G. M., Selles, R. W., van der Geest, J. N., & Donchin, O. (2018) · eNeuro · doi:10.1523/eneuro.0170-18.2018 · Cited by 37
Laplace Approximation for Logistic Gaussian Process Density Estimation and Regression
Riihimäki, J., & Vehtari, A. (2014) · Bayesian Analysis · doi:10.1214/14-ba872 · Cited by 62
Thom, D., & Keeton, W. S. (2020) · Forest Ecology and Management · doi:10.1016/j.foreco.2020.118132 · Cited by 45
Jaya, I. G. N. M., & Folmer, H. (2020) · Geographical Analysis · doi:10.1111/gean.12264 · Cited by 41
Spanbauer, C., & Sparapani, R. (2021) · Statistics in Medicine · doi:10.1002/sim.8924 · Cited by 49
Reliability-oriented sensitivity analysis in presence of data-driven epistemic uncertainty
Sarazin, G., Morio, J., Lagnoux, A., Balesdent, M., & Brevault, L. (2021) · Reliability Engineering & System Safety · doi:10.1016/j.ress.2021.107733 · Cited by 41
Wand, M. P. (2016) · UTS ePRESS (University of Technology Sydney) · doi:10.6084/m9.figshare.3462980.v1 · Cited by 38
Bayesian model selection for logistic regression with misclassified outcomes
Gerlach, R., & Stamey, J. D. (2007) · Statistical Modelling · doi:10.1177/1471082x0700700303 · Cited by 40
“Note that Gelman (2006) suggests a flat prior, restricted to a finite range, for the second level standard deviation σ (not the variance) in a normal hierarchical model, given a sufficient number of groups (say > 5).”
BackgroundPaget, M., Amorós, W., Salas, E., Eyzaguirre, R., Alspach, P. A., Apiolaza, L. A., Noble, A. D. L., & Bonierbale, M. (2014) · Crop Science · doi:10.2135/cropsci2013.12.0809 · Cited by 39
A General Model of Abstention Under Compulsory Voting
Katz, G., & Levin, I. (2016) · Political Science Research and Methods · doi:10.1017/psrm.2016.49 · Cited by 38
“In contrast, recent work (Gelman 2006; Austin 2010) has shown that Bayesian methods yield accurate estimates of the regression parameters and variance components of hierarchical models with very few clusters, provided the number of observations per group is reasonably large—as is the case in our…”
ResultReducing the conservation reliance of the endangered Kirtland's warbler through adaptive management
Cooper, N. W., Rushing, C. S., & Marra, P. P. (2019) · Journal of Wildlife Management · doi:10.1002/jwmg.21714 · Cited by 39
SDPRX: A statistical method for cross-population prediction of complex traits
Zhou, G., Chen, T., & Zhao, H. (2022) · The American Journal of Human Genetics · doi:10.1016/j.ajhg.2022.11.007 · Cited by 38
Finite mixture modeling in time series: A survey of Bayesian filters and fusion approaches
Li, T., Liang, H., Xiao, B., Pan, Q., & He, Y. (2023) · Information Fusion · doi:10.1016/j.inffus.2023.101827 · Cited by 42
A Hierarchical Expected Improvement Method for Bayesian Optimization
Chen, Z., Mak, S., & Wu, C. (2023) · Journal of the American Statistical Association · doi:10.1080/01621459.2023.2210803 · Cited by 70
“Whileweakly informative (and noninformative) priors are widely used in Bayesian analysis (Gelman 2006), we have found that such priors can result in poor optimization performance for HEI (Section 6 provides further details).”
MethodCalibrating and validating bacterial water quality models: A Bayesian approach
Gronewold, A. D., Qian, S. S., Wolpert, R. L., & Reckhow, K. H. (2009) · Water Research · doi:10.1016/j.watres.2009.02.034 · Cited by 38
“…2004, Chapter 4 and Section 4.6, respectively) for tAe and s A2 from equation (10) using Bayesian Markov chain Monte Carlo (MCMC) procedures in the software program WinBUGS (Lunn et al., 2000) with diffuse prior distributions (for details, see Gelman, 2006; Gelman and Hill, 2007, pp. 345–355).”
MethodInfluentialAn application of Harrell's C‐index to PH frailty models
Oirbeek, R. V., & Lesaffre, E. (2010) · Statistics in Medicine · doi:10.1002/sim.4058 · Cited by 45
Cancer mortality inequalities in urban areas: a Bayesian small area analysis in Spanish cities
Puigpinós‐Riera, R., Marí-Dell’Olmo, M., Gotsens, M., Borrell, C., Serral, G., Ascaso, C., Calvo, M., Daponte, A., Domínguez‐Berjón, M. F., Esnaola, S., et al. (2011) · International Journal of Health Geographics · doi:10.1186/1476-072x-10-6 · Cited by 37
Basis Function Models for Animal Movement
Hooten, M. B., & Johnson, D. S. (2016) · Journal of the American Statistical Association · doi:10.1080/01621459.2016.1246250 · Cited by 45
“For example, we used a proper uniformprior forσ , as recommended by Gelman (2006) for latent Gaussian processes.”
MethodUncertain links in host–parasite networks: lessons for parasite transmission in a multi-host system
Walker, J. G., Plein, M., Morgan, E. R., & Vesk, P. A. (2017) · Philosophical Transactions of the Royal Society B Biological Sciences · doi:10.1098/rstb.2016.0095 · Cited by 44
A Bayesian nonlinear random effects model for identification of defective batteries from lot samples
Cripps, E., & PECHT, M. G. (2016) · Journal of Power Sources · doi:10.1016/j.jpowsour.2016.12.067 · Cited by 39
Rethinking safety in numbers: are intersections with more crossing pedestrians really safer?
Xu, P., Xie, S., Dong, N., et al. (2017) · Injury Prevention · doi:10.1136/injuryprev-2017-042469 · Cited by 38
A Mixed-Effects Location Scale Model for Dyadic Interactions
Rast, P., & Ferrer, E. E. (2018) · Multivariate Behavioral Research · doi:10.1080/00273171.2018.1477577 · Cited by 43
“To ensure good quality of the parameter estimates, we chose to keep the number of iterations at a level where the models converged with potential scale reduction factors R̂ smaller than 1.1 (cf. Gelman, 2006).”
BackgroundMethodInfluentialA Bayesian nonlinear mixed-effects location scale model for learning
Williams, D. R., Zimprich, D., & Rast, P. (2019) · Behavior Research Methods · doi:10.3758/s13428-019-01255-9 · Cited by 46
“This number of iterations provided a good quality of the parameter estimates in which the models converged with potential scale reduction factors R̂ smaller than 1.1 (Gelman, 2006).”
InfluentialAssessing the quality of life history information in publicly available databases
Thorson, J. T., Cope, J. M., & Patrick, W. S. (2013) · Ecological Applications · doi:10.1890/12-1855.1 · Cited by 37
Conspicuous visual signals do not coevolve with increased body size in marine sea slugs
Cheney, K. L., Cortesi, F., How, M. J., et al. (2014) · Journal of Evolutionary Biology · doi:10.1111/jeb.12348 · Cited by 38
“For r, we used a Uniform prior on [0, 100] (Gelman, 2006).”
MethodTesting Random Effects in the Linear Mixed Model Using Approximate Bayes Factors
Saville, B. R., & Herring, A. H. (2008) · Biometrics · doi:10.1111/j.1541-0420.2008.01107.x · Cited by 45
“Gelman (2006) discusses various approaches to default priors specifically for variance components.”
BackgroundNarayan, G., Matheson, T., Saha, A., Axelrod, T., Calamida, A., Olszewski, E. W., Claver, J., Mandel, K. S., Bohlin, R. C., Holberg, J. B., et al. (2019) · The Astrophysical Journal Supplement Series · doi:10.3847/1538-4365/ab0557 · Cited by 44
“We follow Gelman (2006) and use weakly informative priors on all the parameters of our hierarchical model.”
MethodA framework to model global, regional, and national estimates of intimate partner violence
Maheu‐Giroux, M., Sardinha, L., Stöckl, H., Meyer, S. R., Godin, A., Alexander, M. J., & Garcia‐Moreno, C. (2022) · BMC Medical Research Methodology · doi:10.1186/s12874-022-01634-5 · Cited by 46
“Vehtari A, Gelman A, Gabry J. Practical Bayesian model evalua‑ tion using leave‑one‑out cross‑validation and WAIC.”
MethodInfluentialIngebrigtsen, R., Lindgren, F., Steinsland, I., & Martino, S. (2015) · Spatial Statistics · doi:10.1016/j.spasta.2015.07.003 · Cited by 46
“Prior specifications for hierarchical model parameters are challenging for applied Bayesians, and common priors might have undesired properties (Gelman, 2006; Sørbye and Rue, 2014; Simpson et al., 2014).”
BackgroundHierarchical Bayesian Inference and Recursive Regularization for Large-Scale Classification
Gopal, S., & Yang, Y. (2015) · ACM Transactions on Knowledge Discovery from Data · doi:10.1145/2629585 · Cited by 37
“Neither worked well, the former being too sensitive to the value of (cid:13) , which is in agreement with the observations made by [Gelman 2006] and the latter constraining the model by using a single a and b .”
ResultGotsens, M., Marí-Dell’Olmo, M., Martínez‐Beneito, M. A., Pérez, K., Pasarín, M. I., Daponte, A., Puigpinós‐Riera, R., Rodríguez‐Sanz, M., Audicana, C., Nolasco, A., et al. (2011) · Accident Analysis & Prevention · doi:10.1016/j.aap.2011.04.013 · Cited by 38
Random projections for Bayesian regression
Geppert, L. N., Ickstadt, K., Munteanu, A., Quedenfeld, J., & Sohler, C. (2015) · Statistics and Computing · doi:10.1007/s11222-015-9608-z · Cited by 59
“We choose an improper uniform prior rather than an inverse gamma prior with small values for the hyperparameters as Gelman (2006) indicates that such priors can have a skewing effect on the posterior distribution.”
MethodSimilar recent selection criteria associated with different behavioural effects in two dog breeds
Sundman, A., Johnsson, M., Wright, D., & Jensen, P. H. K. (2016) · Genes Brain & Behavior · doi:10.1111/gbb.12317 · Cited by 41
“We used a parameter expanded prior (Gelman 2006) for the additive genetic variance component with V =1, 𝜈 = 1, 𝛼𝜇 =0, 𝛼V = 1000, which corresponds to a half-Cauchy prior with scale 100.”
MethodA Reliability‐Based Capability Approach
Tabandeh, A., Gardoni, P., & Murphy, C. (2017) · Risk Analysis · doi:10.1111/risa.12843 · Cited by 42
Blazar, D. (2017) · Education Finance and Policy · doi:10.1162/edfp_a_00251 · Cited by 46
Zhang, X., Xu, C., & Xiao, G. (2018) · BMC Infectious Diseases · doi:10.1186/s12879-018-3546-2 · Cited by 36
Rub, A. M. W., Som, N. A., Henderson, M. J., Sandford, B. P., Doornik, D. M. V., Teel, D. J., Tennis, M. J., Langness, O. P., van der Leeuw, B. K., & Huff, D. D. (2019) · Canadian Journal of Fisheries and Aquatic Sciences · doi:10.1139/cjfas-2018-0290 · Cited by 42
Wang, L., Chen, X., Kang, S., Deng, X., & Jin, R. (2020) · Additive manufacturing · doi:10.1016/j.addma.2020.101211 · Cited by 44
“The traditional Markov chain Monte Carlo (MCMC) samplers, such as Gibbs [44], usually lead to a large fraction of rejected samples.”
BackgroundModelling subject variability in the spatial and temporal characteristics of functional modes
Harrison, S. J., Bijsterbosch, J. D., Segerdahl, A. R., Fitzgibbon, S. P., Farahibozorg, S.-R., Duff, E. P., Smith, S. M., & Woolrich, M. W. (2020) · NeuroImage · doi:10.1016/j.neuroimage.2020.117226 · Cited by 44
Genetic basis of growth, spring phenology, and susceptibility to biotic stressors in maritime pine
Hurel, A., de Miguel, M., Dutech, C., Desprez‐Loustau, M., Plomion, C., Rodríguez‐Quilón, I., Cyrille, A., Guzman, T., Alı́a, R., González‐Martínez, S. C., et al. (2021) · Evolutionary Applications · doi:10.1111/eva.13309 · Cited by 36
“Parameter expanded priors were used to improve the convergence and mixing properties of the chain as suggested by Gelman (2006) for models on the presence of pine processionary moth nests, needle discoloration caused by D. sapinea, and necrosis caused by A. ostoyae.”
MethodSraj, I., Maître, O. P. L., Knio, O., et al. (2015) · Computer Methods in Applied Mechanics and Engineering · doi:10.1016/j.cma.2015.10.002 · Cited by 37
“1 and l max = 1, and an inverse Gamma prior [42, 43, 44] for σ 2 f with parameters α = 3 and β = 1.”
BackgroundAnother Look at Bayesian Analysis of AMMI Models for Genotype-Environment Data
Josse, J., van Eeuwijk, F. A., Piepho, H., & Denis, J. (2014) · Journal of Agricultural Biological and Environmental Statistics · doi:10.1007/s13253-014-0168-z · Cited by 37
Integrated data analysis in the presence of emigration and mark loss
Reynolds, T. J., King, R., Harwood, J., et al. (2009) · Journal of Agricultural Biological and Environmental Statistics · doi:10.1198/jabes.2009.08008 · Cited by 37
Childhood Brain Cancer in Florida: A Bayesian Clustering Approach
Lawson, A. B., & Rotejanaprasert, C. (2014) · Statistics and Public Policy · doi:10.1080/2330443x.2014.970247 · Cited by 37
“Each of the main terms on the BYM model has a Gaussian prior distribution and their precisions are modeled via the stable SD-uniform specification above (Gelman 2006).”
MethodSmall area estimation of the Gini concentration coefficient
Fabrizi, E., & Trivisano, C. (2016) · Computational Statistics & Data Analysis · doi:10.1016/j.csda.2016.01.010 · Cited by 41
Otsu, C., Iijima, H., Nagaike, T., & Hoshino, Y. (2017) · Journal of Vegetation Science · doi:10.1111/jvs.12514 · Cited by 37
“The prior distribution of ea,i was defined as normal (0, r (2)); hyper priors of lAT were normal (0, r (2)); and hyper priors of r were uniform (0, 100) (Gelman 2006).”
BackgroundThe fitness consequences of wildlife conservation translocations: a meta‐analysis
Gross, I. P., Wilson, A. E., & Wolak, M. E. (2023) · Biological reviews/Biological reviews of the Cambridge Philosophical Society · doi:10.1111/brv.13025 · Cited by 45
“…prior distributions for all fi xed effects (mean-= 0, variance = 10 10 ) and parameter expanded priors on all variance components to yield scaled non-central F-distributions with numerator and denominator degrees of freedom equal to 1 and a scale parameter of 1000 (Gelman, 2006; Had fi eld, 2010).”
BackgroundHobbs, B. P., Thall, P. F., & Lin, S. H. (2015) · Journal of the Royal Statistical Society Series C (Applied Statistics) · doi:10.1111/rssc.12117 · Cited by 37
“…effects: δψk ∼ N.δψ,ω2/, k = 1, : : : , 7: A non-informative prior was assumed for the hierarchical mean, δψ ∼ N.0, 100/: Following the recommendations of Gelman (2006), the hierarchical standard deviation ω was assumed to be uniform over .0, 10]: For the treatment effects for surgery, δ̃,…”
BackgroundEvaluating protected area effectiveness using bird lists in the Australian Wet Tropics
Barnes, M. D., Szabo, J. K., Morris, W. K., & Possingham, H. P. (2014) · Diversity and Distributions · doi:10.1111/ddi.12274 · Cited by 41
“Diversity and Distributions, 21, 368–378, ª 2014 John Wiley & Sons Ltd 371 We used vague priors (Gelman, 2006) for the regression coefficients (normal distribution with means zero and standard deviation 10,000).”
MethodVariational Inference for Count Response Semiparametric Regression
Luts, J., & WAND, M. P. (2015) · Bayesian Analysis · doi:10.1214/14-ba932 · Cited by 46
Authier, M., Peltier, H., Dorémus, G., Dabin, W., Canneyt, O. V., & Ridoux, V. (2014) · Biodiversity and Conservation · doi:10.1007/s10531-014-0741-3 · Cited by 43
“Weakly informative priors were used (Gelman 2006; Gelman et al. 2008).”
MethodYan, J., Guo, C., & Paarlberg, L. E. (2014) · The American Statistician · doi:10.1080/00031305.2014.955211 · Cited by 40
“Recently, Gelman (2006) proposed uniform prior distributions for standard deviation parameters with range restricted to reasonable values on the positive domain.”
BackgroundVanhatalo, J. P., Veneranta, L., & Hudd, R. (2012) · Ecological Modelling · doi:10.1016/j.ecolmodel.2011.12.025 · Cited by 50
A Framework for Interpreting Climate Model Outputs
Leith, N., & Chandler, R. E. (2009) · Journal of the Royal Statistical Society Series C (Applied Statistics) · doi:10.1111/j.1467-9876.2009.00694.x · Cited by 45
“This choice of Lφ ensures that, within a reasonable range, the prior does not dominate the data and also safeguards against an unrealistically long right-hand tail in the posterior for σφ, which can be a problem when non-informative uniform priors are used in such settings (Gelman, 2006).”
BackgroundMethodInfluentialChecking for Prior-Data Conflict Using Prior-to-Posterior Divergences
Nott, D. J., Wang, X., Evans, M. J., & Englert, B. (2020) · Statistical Science · doi:10.1214/19-sts731 · Cited by 48
Bayesian analysis of homogeneity studies in the production of reference materials
van der Veen, A. M. H. (2017) · Accreditation and Quality Assurance · doi:10.1007/s00769-017-1292-6 · Cited by 36
Quick, M., Law, J., & Li, G. (2017) · Environment and Planning B Urban Analytics and City Science · doi:10.1177/2399808317744779 · Cited by 36
“The Normalþ1 (0, 10) distribution was independently assigned as the prior for the standard deviations of the random effects si, ui, j, ij, and j (Gelman, 2006).”
MethodEstimating the Relative Excess Risk Due to Interaction in Clustered-Data Settings
Correia, K. F. B., & Williams, P. L. (2018) · American Journal of Epidemiology · doi:10.1093/aje/kwy154 · Cited by 45
Using a hidden Markov model to measure earnings quality
Du, K., Huddart, S. J., Xue, L., & Zhang, Y. (2019) · Journal of Accounting and Economics · doi:10.1016/j.jacceco.2019.101281 · Cited by 39
“We refer readers to the Bayesian analysis literature prior research For more rigorous discussions of the technical advantages of the Bayesian hierarchical framework (Gelman, 2006; Gelman et al., 2013).”
RoBoT: a robust Bayesian hypothesis testing method for basket trials
Zhou, T., & Ji, Y. (2020) · Biostatistics · doi:10.1093/biostatistics/kxaa005 · Cited by 36
Hierarchical Bayesian approaches to statistical modelling of geotechnical data
Bozorgzadeh, N., & Bathurst, R. J. (2020) · Georisk Assessment and Management of Risk for Engineered Systems and Geohazards · doi:10.1080/17499518.2020.1864411 · Cited by 53
“key aspects of hierarchical Bayesian modelling as discussed in seminal Bayesian references such as Gelman et al. (2013) and Lunn et al.”
BackgroundMethodResultInfluentialPrecipitation reconstruction from climate-sensitive lithologies using Bayesian machine learning
Chandra, R., Cripps, S., Butterworth, N. P., & Müller, R. D. (2021) · Environmental Modelling & Software · doi:10.1016/j.envsoft.2021.105002 · Cited by 33
Hupf, B., Bunn, V., Lin, J., & Dong, C. (2021) · Statistics in Medicine · doi:10.1002/sim.8970 · Cited by 45
Estimation of covariance matrices based on hierarchical inverse-Wishart priors
Bouriga, M., & Féron, O. (2012) · Journal of Statistical Planning and Inference · doi:10.1016/j.jspi.2012.09.006 · Cited by 36
Relative magnitude of cohort, age, and year effects on size at age of exploited marine fishes
Thorson, J. T., & Minte‐Vera, C. V. (2014) · Fisheries Research · doi:10.1016/j.fishres.2014.11.016 · Cited by 43
Fitch, D. T., Rhemtulla, M., & Handy, S. (2018) · Transportation Research Part A Policy and Practice · doi:10.1016/j.tra.2018.06.013 · Cited by 50
Bayesian hierarchical models for network meta-analysis incorporating nonignorable missingness
Zhang, J., Chu, H., Hong, H., Virnig, B. A., & Carlin, B. P. (2015) · Statistical Methods in Medical Research · doi:10.1177/0962280215596185 · Cited by 34
Transformational party events and legislative turnover in West European democracies, 1945–2015
Gouglas, A., Katz, G., Maddens, B., & Brans, M. (2020) · Party Politics · doi:10.1177/1354068820944703 · Cited by 37
“By contrast, recent work (e.g. Gelman, 2006) has shown that Bayesian inferential methods such as those proposed by Galvis et al. (2014) to fit the augmented beta regression model yield accurate estimates for hierarchical models with fewer than 10 clusters.”
ResultBull, L. A., Francesco, D. D., Dhada, M. H., Steinert, O., Lindgren, T., Parlikad, A. K., Duncan, A., & Girolami, M. (2022) · Computer-Aided Civil and Infrastructure Engineering · doi:10.1111/mice.12901 · Cited by 48
Alameddine, I., Qian, S., & Reckhow, K. (2011) · Water Research · doi:10.1016/j.watres.2010.08.003 · Cited by 38
Bellera, C. A., Hanley, J. A., Joseph, L., et al. (2008) · Annals of Epidemiology · doi:10.1016/j.annepidem.2007.10.006 · Cited by 35
Functional Autoregression for Sparsely Sampled Data
Kowal, D. R., Matteson, D. S., & Ruppert, D. (2017) · Journal of Business and Economic Statistics · doi:10.1080/07350015.2017.1279058 · Cited by 53
“Following Gelman (2006), we place uniform priors on the standard deviations λ −1/2 φ j ∼ Uniform(0, 104), which implies the prior for the precision [λφ j ] ∝ λ−3/2φ j 1{λφ j > 10−8}.”
MethodInfluentialRen, S., Oakley, J. E., & Stevens, J. (2018) · Medical Decision Making · doi:10.1177/0272989x18759488 · Cited by 38
Developmental effects of heatwave conditions on the early life stages of a coral reef fish
Spinks, R. K., Munday, P. L., & Donelson, J. M. (2019) · Journal of Experimental Biology · doi:10.1242/jeb.202713 · Cited by 47
“A half Cauchy distribution for the error standard deviation prior was selected as it is ideal with weakly informative priors and hierarchical models (Gelman, 2006).”
MethodMarcondes, R. S., Nations, J. A., Seeholzer, G. F., & Brumfield, R. T. (2021) · The American Naturalist · doi:10.1086/713386 · Cited by 41
Evaluating recent methods to overcome spatial confounding
Urdangarin, A., Goicoa, T., & Ugarte, M. D. (2022) · Revista Matemática Complutense · doi:10.1007/s13163-022-00449-8 · Cited by 43
“Recently, Gilbert et al. (2022) state that spatial confounding is seldom defined explicitly and they point to four phenomena related to this concept.”
BackgroundMethod
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.