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 101–200 of 4,089
Performing Arm-Based Network Meta-Analysis in R with the pcnetmeta Package
Lin, L., Zhang, J., Hodges, J. S., & Chu, H. (2017) · Journal of Statistical Software · doi:10.18637/jss.v080.i05 · Cited by 195
“However, the posterior may be sensitive to the choice of ε , and a uniform prior for standard deviation is preferred in some cases (Gelman 2006).”
BackgroundProbabilistic Common Spatial Patterns for Multichannel EEG Analysis
Wu, W., Chen, Z., Gao, X., Li, Y., Brown, E. N., & Gao, S. (2015) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2014.2330598 · Cited by 194
Modeling between-trial variance structure in mixed treatment comparisons
Lu, G., & Ades, A. (2009) · Biostatistics · doi:10.1093/biostatistics/kxp032 · Cited by 196
“Comments on the weakly informative specification for the hyperparameters can be referred to Warn and others (2002), Spiegalhalter and others (2004), and Gelman (2006). An apparent feature of the treatment regime in GERD data is that both therapies H2RA and PPI were used in single and double doses, which are regarded a…”
BackgroundCosts of multimorbidity: a systematic review and meta-analyses
Tran, P. B., Kazibwe, J., Nikolaidis, G. F., Linnosmaa, I., Rijken, M., & van Olmen, J. (2022) · BMC Medicine · doi:10.1186/s12916-022-02427-9 · Cited by 192
Nalborczyk, L., Batailler, C., Lœvenbruck, H., Vilain, A., & Bürkner, P.-C. (2019) · Journal of Speech, Language, and Hearing Research · doi:10.1044/2018_jslhr-s-18-0006 · Cited by 183
Educational Differences in U.S. Adult Mortality: A Cohort Perspective
Masters, R. K., Hummer, R. A., & Powers, D. A. (2012) · American Sociological Review · doi:10.1177/0003122412451019 · Cited by 188
“We assumed noninformative prior distributions for all model parameters (Gelman 2006; Lynch 2007).”
MethodThomson, J. R., Kimmerer, W. J., Brown, L. R., Newman, K. B., Nally, R. M., Bennett, W. A., Feyrer, F., & Fleishman, E. (2010) · Ecological Applications · doi:10.1890/09-0998.1 · Cited by 187
Sorensen, T., Hohenstein, S., & Vasishth, S. (2016) · The Quantitative Methods for Psychology · doi:10.20982/tqmp.12.3.p175 · Cited by 180
“For those looking for a psychology-specific introduction, the books by Kruschke (2014) and Lee and Wagenmakers (2013) are to be recommended, although for the latter the going might be easier if the reader has already looked at Gelman and Hill (2007). As a second book, David Lunn et al. (2012) is recommended; it provid…”
BackgroundMethodInfluentialDisturbance type and species life history predict mammal responses to humans
Suraci, J. P., Gaynor, K. M., Allen, M. L., Alexander, P., Brashares, J. S., Cendejas‐Zarelli, S., Crooks, K., Elbroch, L. M., Forrester, T., Green, A. M., et al. (2021) · Global Change Biology · doi:10.1111/gcb.15650 · Cited by 186
“We confirmed con - vergence of HMC chains by visual inspection of trace plots and via the Gelman– Rubin statistic (“ R - hat” values <1.1 for all parameters; Gelman, 2006) and verified the reliability of model inference follow - ing the procedure outlined by Gabry et al. (2019).”
MethodInfluentialRuscoe, W. A., Ramsey, D. S. L., Pech, R. P., Sweetapple, P. J., Yockney, I., Barron, M. C., Perry, M., Nugent, G., Carran, R., Warne, R., et al. (2011) · Ecology Letters · doi:10.1111/j.1461-0248.2011.01673.x · Cited by 187
General Design Bayesian Generalized Linear Mixed Models
Zhao, Y., Staudenmayer, J., Coull, B. A., & Wand, M. P. (2006) · Statistical Science · doi:10.1214/088342306000000015 · Cited by 183
“For hierarchical versions of GLMMs, however, recent research has shown that these priors can actually be quite informative, leading to inferences that are sensitive to choice of the hyperparameters for these distributions (Natarajan and McCulloch 1998; Natarajan and Kass 2000; Gelman 2004). Natarajan and Kass (1998) a…”
BackgroundInfluentialMeta‐analysis of few small studies in orphan diseases
Friede, T., Röver, C., Wandel, S., & Neuenschwander, B. (2017) · Research Synthesis Methods · doi:10.1002/jrsm.1217 · Cited by 186
“These weakly informative priors can be represented by half-normal, half-Cauchy or half-t distributions (Spiegelhalter et al., 2004; Gelman, 2006; Polson and Scott, 2012).”
BackgroundZucker, D. R., Ruthazer, R., & Schmid, C. H. (2010) · Journal of Clinical Epidemiology · doi:10.1016/j.jclinepi.2010.04.020 · Cited by 184
Gallo, T., Fidino, M., Lehrer, E. W., & Magle, S. B. (2017) · Ecological Applications · doi:10.1002/eap.1611 · Cited by 181
“We assumed that rs follows a semi-informed Inv–Gamma(3,2) hyperdistribution (Gelman 2006).”
BackgroundThree case studies in the Bayesian analysis of cognitive models
LEE, M. D. (2008) · Psychonomic Bulletin & Review · doi:10.3758/pbr.15.1.1 · Cited by 177
“For example, (3)Gelman (2006) points out that this prior distribution, although very widely used, is problematic for small variances.”
BackgroundRöver, C., Bender, R., Dias, S., Schmid, C. H., Schmidli, H., Sturtz, S., Weber, S., & Friede, T. (2021) · Research Synthesis Methods · doi:10.1002/jrsm.1475 · Cited by 179
“In such a case, an unequal pair of two data points for example implies that the variance must be positive (a zero variance would have a zero likelihood); in the present NNHM context, however, unequal yi values may be consistent with zero heterogeneity (τ = 0), so that such priors are not a natural choice here, and the…”
BackgroundInfluentialHuman metabolic profiles are stably controlled by genetic and environmental variation
Nicholson, G., Rantalainen, M., Maher, A. D., Li, J. V., Malmodin, D., Ahmadi, K. R., Faber, J. H., Hallgrímsdóttir, I. B., Barrett, A., Toft, H., et al. (2011) · Molecular Systems Biology · doi:10.1038/msb.2011.57 · Cited by 176
“…corresponding variance parameter (denoted by su 2), was defined to be: pðujs2uÞ qðujm ¼ 0;s2 ¼ s2u; n ¼ 1; d ¼ 0:1Þ: Independent Uniform priors were placed on the standard deviation parameters (Gelman, 2006), p(s)BUniform(s|0, 10 sy), where sy denotes the sample standard deviation of the data, y.”
Robust exchangeability designs for early phase clinical trials with multiple strata
Neuenschwander, B., Wandel, S., Roychoudhury, S., & Bailey, S. (2016) · Pharmaceutical Statistics · doi:10.1002/pst.1730 · Cited by 175
Bias and Precision in Q ST Estimates: Problems and Some Solutions
O'Hara, R. B., & Merilä, J. (2005) · Genetics · doi:10.1534/genetics.105.044545 · Cited by 176
Boggs, A. M., Wali, B., & Khattak, A. J. (2020) · Accident Analysis & Prevention · doi:10.1016/j.aap.2019.105354 · Cited by 176
Kogan, N. E., Clemente, L., Liautaud, P., Kaashoek, J., Link, N. B., Nguyen, A. T., Lu, F. S., Huybers, P., Resch, B., Havas, C., et al. (2021) · Science Advances · doi:10.1126/sciadv.abd6989 · Cited by 172
Baribault, T. W., Kobe, R. K., & Finley, A. O. (2012) · Ecological Monographs · doi:10.1890/11-1013.1 · Cited by 175
Responses to monetary policy shocks in the east and the west of Europe: a comparison
Jarociński, M. (2010) · Journal of Applied Econometrics · doi:10.1002/jae.1082 · Cited by 193
Using a Bayesian hierarchical model to improve Lake Erie cyanobacteria bloom forecasts
Obenour, D. R., Gronewold, A. D., Stow, C. A., & Scavia, D. (2014) · Water Resources Research · doi:10.1002/2014wr015616 · Cited by 175
Human evaluation of automatically generated text: Current trends and best practice guidelines
van der Lee, C., Gatt, A., van Miltenburg, E., & Krahmer, E. (2021) · Computer Speech & Language · doi:10.1016/j.csl.2020.101151 · Cited by 170
Spatial and temporal drivers of avian population dynamics across the annual cycle
Rushing, C. S., Hostetler, J. A., Sillett, T. S., Marra, P. P., Rotenberg, J. A., & Ryder, T. B. (2017) · Ecology · doi:10.1002/ecy.1967 · Cited by 171
“We used half-Cauchy priors with scale parameters = 10 and 15 for the annual and monthly variance parameters, respectively (Gelman 2006).”
MethodMultiple Imputation for Multilevel Data with Continuous and Binary Variables
Audigier, V., White, I. R., Jolani, S., Debray, T. P. A., Quartagno, M., Carpenter, J., van Buuren, S., & Resche-Rigon, M. (2018) · Statistical Science · doi:10.1214/18-sts646 · Cited by 166
Duncan, R. P., Clemants, S. E., Corlett, R. T., Hahs, A. K., McCarthy, M. A., McDonnell, M. J., Schwartz, M. W., Thompson, K., Vesk, P. A., & Williams, N. S. G. (2011) · Global Ecology and Biogeography · doi:10.1111/j.1466-8238.2010.00633.x · Cited by 165
Rising, J., & Devineni, N. (2020) · Nature Communications · doi:10.1038/s41467-020-18725-w · Cited by 165
Model adequacy and the macroevolution of angiosperm functional traits
Pennell, M. W., FitzJohn, R. G., Cornwell, W. K., & Harmon, L. J. (2015) · The American Naturalist · doi:10.1086/682022 · Cited by 165
Ching, J., & Phoon, K.-K. (2019) · Journal of Engineering Mechanics · doi:10.1061/(asce)em.1943-7889.0001537 · Cited by 164
Depaoli, S., & Clifton, J. P. (2015) · Structural Equation Modeling: A Multidisciplinary Journal · doi:10.1080/10705511.2014.937849 · Cited by 163
“The tendency for priors to have a large impact on estimates for categorical variables is a recognized phenomenon in the SEM literature that has been referred to as prior assumption dependence (Asparouhov & Muthén, 2010).”
BackgroundResultInfluentialMeadows, A., Kaambwa, B., Novielli, N., Huissoon, A., Fry-Smith, A., Meads, C., Barton, P., & Dretzke, J. (2013) · Health Technol Assess · doi:10.3310/hta17270 · Cited by 163
Sensitivity analysis in Bayesian generalized linear mixed models for binary data
Roos, M., & Held, L. (2011) · Bayesian Analysis · doi:10.1214/11-ba609 · Cited by 159
“Consequently, we agree with Gelman (2006) and Lunn et al. (2009b) that one should be careful with gamma priors.”
BackgroundMethodInfluentialA Genomic Bayesian Multi-trait and Multi-environment Model
Montesinos-López, O. A., Montesinos-López, A., Crossa, J., Toledo, F. H., Pérez-Hernández, O., Eskridge, K. M., & Rutkoski, J. (2016) · G3 Genes|Genomes|Genetics · doi:10.1534/g3.116.032359 · Cited by 159
“However, Gelman (2006) and Huang and Wand (2013) argued against using IW priors for covariance matrices because they impose a degree of informativity and the posterior inferences are sensitive to the choice of hyper-parameters.”
MethodHierarchical Bayesian continuous time dynamic modeling.
Driver, C. C., & Voelkle, M. C. (2018) · Psychological Methods · doi:10.1037/met0000168 · Cited by 157
The Horseshoe+ Estimator of Ultra-Sparse Signals
Bhadra, A., Datta, J., Polson, N. G., & Willard, B. T. (2016) · Bayesian Analysis · doi:10.1214/16-ba1028 · Cited by 158
“Johnstone and Silverman (2004) showed that a thresholding-based estimator for θ under the two-groups model with an empirical Bayes estimate for μ is minimax in `2 sense. 2. Castillo and van der Vaart (2012) treated a full Bayes version of the problem and again found an estimate that is minimax in `2.”
BackgroundInfluentialChen, Y.-L., Senande, L. L., Thorsen, M., & Patten, K. (2021) · Autism · doi:10.1177/13623613211005918 · Cited by 161
“…for logistic regression in the literature, including a Student t distribution with 7 degrees of freedom for regression coefficients and intercepts, as well as a halfCauchy prior with 4 degrees of freedom for the variance components (Gelman, 2006; Ghosh et al., 2018; McNeish & Stapleton, 2016).”
MethodTropical snake diversity collapses after widespread amphibian loss
Zipkin, E. F., DiRenzo, G. V., Ray, J. M., Rossman, S., & Lips, K. R. (2020) · Science · doi:10.1126/science.aay5733 · Cited by 192
“fitnew) using only observed species (1-36) p.”
Independent contrasts and PGLS regression estimators are equivalent
Blomberg, S. P., Lefevre, J. G., Wells, J. A., & Waterhouse, M. (2012) · Systematic Biology · doi:10.1093/sysbio/syr118 · Cited by 158
“A uniform prior can be used for σ (Gelman 2006).”
BackgroundMethodStatistical and Clinical Aspects of Hospital Outcomes Profiling
Normand, S.-L. T., & Shahian, D. M. (2007) · Statistical Science · doi:10.1214/088342307000000096 · Cited by 158
BAMLSS: Bayesian Additive Models for Location, Scale, and Shape (and Beyond)
Umlauf, N., Klein, N., & Zeileis, A. (2017) · Journal of Computational and Graphical Statistics · doi:10.1080/10618600.2017.1407325 · Cited by 186
“Therefore, Gelman (2006) proposed the half-Cauchy prior which has the desirable property that for τ ljk = 0 the density is a nonzero constant, whereas the density of the inverse gamma for τ ljk → 0 vanishes (see also Polson and Scott 2012).”
BackgroundBehavioural and neural evidence for self-reinforcing expectancy effects on pain
Jepma, M., Koban, L., van Doorn, J., Jones, M., & Wager, T. D. (2018) · Nature Human Behaviour · doi:10.1038/s41562-018-0455-8 · Cited by 155
Generalized Multilevel Functional Regression
Crainiceanu, C. M., Staicu, A., & Di, C. (2009) · Journal of the American Statistical Association · doi:10.1198/jasa.2009.tm08564 · Cited by 176
“Alternatives to gamma priors are discussed by, for example, Gelman (2006) and Natarajan and Kass (2000), and have the advantage of requiring less care in the choice of the hyperparameters.”
BackgroundMethodMean Field Variational Bayes for Elaborate Distributions
WAND, M. P., Ormerod, J. T., Padoan, S. A., & Fruhwirth, R. K. (2011) · Bayesian Analysis · doi:10.1214/11-ba631 · Cited by 219
Linking fishing mortality reference points to life history traits: an empirical study
Zhou, S., Yin, S., Thorson, J. T., et al. (2012) · Canadian Journal of Fisheries and Aquatic Sciences · doi:10.1139/f2012-060 · Cited by 153
Flexible parametric models for random‐effects distributions
Lee, K. J., & Thompson, S. G. (2008) · Statistics in Medicine · doi:10.1002/sim.2897 · Cited by 148
Models for Potentially Biased Evidence in Meta-Analysis using Empirically Based Priors
Welton, N. J., Ades, A. E., Carlin, J. B., Altman, D. G., & Sterne, J. A. C. (2009) · Journal of the Royal Statistical Society Series A: Statistics in Society · doi:10.1111/j.1467-985x.2008.00548.x · Cited by 150
“Following Gelman et al. (2003) (section 15.”
MethodAnomalously warm temperatures are associated with increased injury deaths
Parks, R. M., Bennett, J. E., Tamura-Wicks, H., Kontis, V., Toumi, R., Danaei, G., & Ezzati, M. (2020) · Nature Medicine · doi:10.1038/s41591-019-0721-y · Cited by 150
Hahn, P. R., Murray, J. S., & Carvalho, C. M. (2017) · SSRN Electronic Journal · doi:10.2139/ssrn.3048177 · Cited by 140
Lima, C., & Lall, U. (2010) · Journal of Hydrology · doi:10.1016/j.jhydrol.2009.12.045 · Cited by 163
Watters, G. M., Hinke, J. T., & Reiss, C. S. (2020) · Scientific Reports · doi:10.1038/s41598-020-59223-9 · Cited by 164
Comparison of Hierarchical Bayesian Models for Overdispersed Count Data using DIC and Bayes' Factors
Millar, R. B. (2009) · Biometrics · doi:10.1111/j.1541-0420.2008.01162.x · Cited by 146
“A half standard-Cauchy prior was chosen for σ and τ (Gelman, 2006).”
MethodPrior Distributions for Objective Bayesian Analysis
Consonni, G., Fouskakis, D., Liseo, B., & Ntzoufras, I. (2018) · Bayesian Analysis · doi:10.1214/18-ba1103 · Cited by 137
Neelon, B., O’Malley, A. J., & Normand, S. T. (2010) · Statistical Modelling · doi:10.1177/1471082x0901000404 · Cited by 154
“However, this choice can be deceptively informative because it concentrates most of the prior mass near zero, which, as Gelman (2006) notes, can lead to unstable posterior estimates since the posterior becomes improper as → 0.”
BackgroundMethodDefault Prior Distributions and Efficient Posterior Computation in Bayesian Factor Analysis
Ghosh, J., & Dunson, D. B. (2009) · Journal of Computational and Graphical Statistics · doi:10.1198/jcgs.2009.07145 · Cited by 142
Devleesschauwer, B., Haagsma, J. A., Angulo, F. J., Bellinger, D. C., Cole, D., Döpfer, D., Fazil, A., Fèvre, E. M., Gibb, H. J., Hald, T., et al. (2015) · PLOS ONE · doi:10.1371/journal.pone.0142498 · Cited by 148
Mixed and Mixture Regression Models for Continuous Bounded Responses Using the Beta Distribution
Verkuilen, J., & Smithson, M. J. (2011) · Journal of Educational and Behavioral Statistics · doi:10.3102/1076998610396895 · Cited by 146
“We found very little difference between results using either of these priors in the examples presented here, but both Browne and Draper (2006) and Gelman (2006) provide useful criticisms and alternatives regarding the effects of using these priors for small sample sizes.”
ResultDisease implications of animal social network structure: A synthesis across social systems
Sah, P., Mann, J., & Bansal, S. (2017) · Journal of Animal Ecology · doi:10.1111/1365-2656.12786 · Cited by 162
Bayesian Analysis of Tests with Unknown Specificity and Sensitivity
(2020) · doi:10.1111/rssc.12435 · Cited by 129
Self-citationEmslie, M. J., Logan, M., Williamson, D. H., Ayling, A. M., MacNeil, M. A., Ceccarelli, D. M., Cheal, A. J., Evans, R. D., Johns, K. A., Jonker, M. J., et al. (2015) · Current Biology · doi:10.1016/j.cub.2015.01.073 · Cited by 166
Li, B., Lingsma, H. F., Steyerberg, E. W., & Lesaffre, E. (2011) · BMC Medical Research Methodology · doi:10.1186/1471-2288-11-77 · Cited by 142
Predicting travel time reliability using mobile phone GPS data
Woodard, D. B., Nogin, G., Koch, P., Racz, D., Goldszmidt, M., & Horvitz, E. (2016) · Transportation Research Part C Emerging Technologies · doi:10.1016/j.trc.2016.10.011 · Cited by 145
“…on unbounded parameter spaces are commonly used in situations where there is little or no prior information regarding the parameter values (Section 2.8 of Gelman et al. (2013)); see for instance Gelman (2006) for the use of uniform priors for standard deviation parameters like τ and σ j,b,q .”
InfluentialIncluding Source Uncertainty and Prior Information in the Analysis of Stable Isotope Mixing Models
Ward, E. J., Semmens, B. X., & Schindler, D. E. (2010) · Environmental Science & Technology · doi:10.1021/es100053v · Cited by 141
Kwon, H., Brown, C., & Lall, U. (2008) · Geophysical Research Letters · doi:10.1029/2007gl032220 · Cited by 147
“The software employs the Gibbs sampler, a Markov Chain Monte Carlo (MCMC) method for simulating the posterior probability distribution of the data field conditional on the current choice of parameters [Gelman, 2006].”
MethodInfluentialFlexible Bayesian quantile regression for independent and clustered data
Reich, B. J., Bondell, H. D., & WANG, H. J. (2009) · Biostatistics · doi:10.1093/biostatistics/kxp049 · Cited by 141
“The standard deviationsσ1k, σ2k ∼ Uniform(0,c1) for some large constantc1 (Gelman, 2006).”
BackgroundBlack-swan events in animal populations
Anderson, S. C., Branch, T. A., Cooper, A. B., & Dulvy, N. K. (2017) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1611525114 · Cited by 161
Statistical and Clinical Aspects of Hospital Outcomes ProfilingPreprint
Normand, S. T., & Shahian, D. M. (2007) · Cited by 141
Tsang, B. L., Devine, O. J., Cordero, A. M., Marchetta, C., Mulinare, J., Mersereau, P. W., Guo, J., Qi, Y., Berry, R. J., Rosenthal, J., et al. (2015) · American Journal of Clinical Nutrition · doi:10.3945/ajcn.114.099994 · Cited by 169
Legacy effects of drought in the southwestern United States: A multi‐species synthesis
Peltier, D. M. P., Fell, M., & Ogle, K. (2016) · Ecological Monographs · doi:10.1002/ecm.1219 · Cited by 148
“6) were assigned uniform, U(0, A), priors, with A = 5 and A = 1, respectively (Gelman 2006).”
Yang, Z.-R., Jiang, Y.-W., Li, F.-X., Liu, D., Lin, T.-F., Zhao, Z.-Y., Wei, C., Jin, Q.-Y., Li, X.-M., Jia, Y.-X., et al. (2023) · The Lancet Microbe · doi:10.1016/s2666-5247(22)00390-1 · Cited by 138
Nicholson, G., Rantalainen, M., Li, J. V., Maher, A. D., Malmodin, D., Ahmadi, K. R., Faber, J. H., Barrett, A., Min, J. L., Rayner, N. W., et al. (2011) · PLoS Genetics · doi:10.1371/journal.pgen.1002270 · Cited by 137
Behmanesh, I., & Moaveni, B. (2016) · Journal of Sound and Vibration · doi:10.1016/j.jsv.2016.03.022 · Cited by 144
Predicting cryptocurrency price bubbles using social media data and epidemic modelling
Phillips, R. C., & Gorse, D. (2017) · IEEE Symposium Series on Computational Intelligence (SSCI) · doi:10.1109/ssci.2017.8280809 · Cited by 138
“Uniform distributions are chosen for the standard deviations, as suggested by Gelman [18]: In the above a and b are hyper-parameters.”
MethodTrends and cycles in economic time series: A Bayesian approach
Harvey, A. C., Trimbur, T. M., & van Dijk, H. K. (2006) · Journal of Econometrics · doi:10.1016/j.jeconom.2006.07.006 · Cited by 124
Prior specification in Bayesian statistics: Three cautionary tales
Dongen, S. V. (2006) · Journal of Theoretical Biology · doi:10.1016/j.jtbi.2006.02.002 · Cited by 134
Genetic constraints predict evolutionary divergence in Dalechampia blossoms
Bolstad, G. H., Hansen, T. F., Pélabon, C., Falahati‐Anbaran, M., Pérez‐Barrales, R., & Armbruster, W. S. (2014) · Philosophical Transactions of the Royal Society B Biological Sciences · doi:10.1098/rstb.2013.0255 · Cited by 143
The Stock Market in the Driver's Seat! Implications for R&D and Marketing
Chakravarty, A., & Grewal, R. (2011) · Management Science · doi:10.1287/mnsc.1110.1317 · Cited by 137
A Bayesian missing data framework for generalized multiple outcome mixed treatment comparisons
Hong, H., Chu, H., Zhang, J., & Carlin, B. P. (2016) · Research Synthesis Methods · doi:10.1002/jrsm.1153 · Cited by 132
“Lambert et al. (2005) and Gelman (2006) caution that Wishart and inverse-gamma priors on variance– covariance matrices can sometimes perform poorly and have computational problems.”
BackgroundMethodInfluentialPrior sensitivity analysis in default Bayesian structural equation modeling.
van Erp, S., Mulder, J., & Oberski, D. L. (2017) · Psychological Methods · doi:10.1037/met0000162 · Cited by 149
“These variances were manipulated because previous research indicates that the vague proper priors for variances are especially influential when the variance parameter is estimated to be close to zero (Gelman, 2006).”
BackgroundMethodResultInfluentialSwart, J. C., Froböse, M. I., Cook, J. L., Geurts, D. E. M., Frank, M. J., Cools, R., & den Ouden, H. E. M. (2017) · eLife · doi:10.7554/elife.22169 · Cited by 152
“The hyperpriors for s are specified by a half-Cauchy (Gelman, 2006) with a scale of 2.”
BackgroundMorel, P., Ulbrich, P., & Gail, A. (2017) · PLoS Biology · doi:10.1371/journal.pbio.2001323 · Cited by 140
Dopaminergic modulation of the exploration/exploitation trade-off in human decision-making
Chakroun, K., Mathar, D., Wiehler, A., Ganzer, F., & Peters, J. (2020) · eLife · doi:10.7554/elife.51260 · Cited by 128
Key Factors Affecting Temporal Variability in Stream Water Quality
Guo, D., Lintern, A. M., Webb, J. A., Ryu, D., Liu, S., Bende‐Michl, U., Leahy, P., Wilson, P., & Western, A. W. (2018) · Water Resources Research · doi:10.1029/2018wr023370 · Cited by 143
“Constituent standard deviation (σ) was assumed to be drawn from a minimally informative zero-truncated normal distribution with mean of 0 and standard deviation of 10 (Gelman, 2006; Stan Development Team, 2018).”
BackgroundFit for a Bayesian: An Evaluation of PPP and DIC for Structural Equation Modeling
Cain, M. K., & Zhang, Z. J. (2018) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705511.2018.1490648 · Cited by 150
“well known that prior choice can affect parameter estimates and substantive conclusions (Gelman, 2006; Gelman & Shalizi, 2013; Johnson, 2013; Seaman, Seaman, & Stamey, 2012; van de Schoot & Depaoli, 2014), but it is less clear how sensitive PPP and DIC are to prior choice.”
BackgroundPark, S. W., Bolker, B. M., Champredon, D., Earn, D. J. D., Li, M., Weitz, J. S., Grenfell, B. T., & Dushoff, J. G. (2020) · Journal of The Royal Society Interface · doi:10.1098/rsif.2020.0144 · Cited by 140
Understanding the Changes in Global Crop Yields Through Changes in Climate and Technology
Najafi, E., Devineni, N., Khanbilvardi, R. M., & Kogan, F. N. (2018) · Earth s Future · doi:10.1002/2017ef000690 · Cited by 150
Spatiotemporal evolution of global population ageing from 1960 to 2017
Li, J., Han, X., Zhang, X., & Wang, S. (2019) · BMC Public Health · doi:10.1186/s12889-019-6465-2 · Cited by 149
“The prior distributions of all random variables in the model are determined as a strictly positive half-Gaussian distribution [22], N+∞(0, 10).”
BackgroundJones, K., Johnston, R. J., Manley, D., Owen, D., & Charlton, C. (2015) · Demography · doi:10.1007/s13524-015-0430-1 · Cited by 136
“Gelman (2006) found that the choice of priors for the variance terms is of the greatest importance when there are few higher-level units and the variance is likely to be small.”
BackgroundPolson, N. G., Scott, J. G., & Windle, J. (2013) · Journal of the Royal Statistical Society Series B (Statistical Methodology) · doi:10.1111/rssb.12042 · Cited by 165
Learning from failures: testing broad taxonomic hypotheses about plant naturalization
Diez, J. M., Williams, P. A., Randall, R. P., Sullivan, J. J., Hulme, P. E., & Duncan, R. P. (2009) · Ecology Letters · doi:10.1111/j.1461-0248.2009.01376.x · Cited by 128
“The overall intercept a0 and regression coefficients were assigned Normal prior distributions with mean 0 and variance 1000, and variance terms at each level are given broad uniform priors on the standard deviation following Gelman (2006).”
MethodSpace–time variability in burglary risk: A Bayesian spatio-temporal modelling approach
Li, G., Haining, R., Richardson, S. T., & Best, N. (2014) · Spatial Statistics · doi:10.1016/j.spasta.2014.03.006 · Cited by 133
Klein, N., Kneib, T., & Lang, S. (2014) · Journal of the American Statistical Association · doi:10.1080/01621459.2014.912955 · Cited by 148
“Half-Cauchy priors have recently been advocated by Polson and Scott (2012); see also Gelman (2006) for a general discussion of prior specifications in Bayesian hierarchical mixed models.”
MethodKuijper, D. P. J., Bubnicki, J. W., Churski, M., Mols, B., & van Hooft, P. (2015) · Behavioral Ecology · doi:10.1093/beheco/arv107 · Cited by 124
Boehm, U., Annis, J. S., Frank, M. J., Hawkins, G. E., Heathcote, A., Kellen, D., Krypotos, A., Lerche, V., Logan, G. D., Palmeri, T. J., et al. (2018) · Journal of Mathematical Psychology · doi:10.1016/j.jmp.2018.09.004 · Cited by 135
“Priors on the standard deviations were weakly informative (Gelman, 2006) half-Cauchy distributions with location parameter 0 and scale parameter 5.”
BackgroundPriors in Bayesian Deep Learning: A Review
Fortuin, V. (2022) · International Statistical Review · doi:10.1111/insr.12502 · Cited by 255
“This means, for instance, that the seemingly innocuous standard Gaussian prior is not uninformative at all [23], but forces the posterior mass onto a thin spherical subspace, which in most cases does not reflect any useful prior knowledge and can severely bias the inference [24, 25].”
BackgroundPlavén-Sigray, P., Matheson, G. J., Collste, K., Ashok, A. H., Coughlin, J. M., Howes, O. D., Mizrahi, R., Pomper, M. G., Rusjan, P., Veronese, M., et al. (2018) · Biological Psychiatry · doi:10.1016/j.biopsych.2018.02.1171 · Cited by 123
“These weakly informative priors were chosen because the numbers of genotype groups (n = 2) and studies (n = 5) are small (40).”
BackgroundLiu, J., Barrett, J. S., Leonardi, E. T., Lee, L., Roychoudhury, S., Chen, Y., & Trifillis, P. (2022) · The Journal of Clinical Pharmacology · doi:10.1002/jcph.2134 · Cited by 175
Goldstein, H., Carpenter, J. R., & Browne, W. J. (2014) · Journal of the Royal Statistical Society Series A: Statistics in Society · doi:10.1111/rssa.12022 · Cited by 121
“…0:001/, p.σ2e /∼Γ−1.0:001, 0:001/: In the literature there are several commonly used diffuse priors for such models (see Browne and Draper (2006) and Gelman (2006) for discussion on this) but this is not of prime interest in this paper and the data set is sufficiently large for the…”
MethodInfluentialGoing Beyond Convergence in Bayesian Estimation: Why Precision Matters Too and How to Assess It
Zitzmann, S., & Hecht, M. (2019) · Structural Equation Modeling: A Multidisciplinary Journal · doi:10.1080/10705511.2018.1545232 · Cited by 122
“…in extensive simulation studies, and have made recommendations for researchers, particularly when sample sizes are small and the estimation of variances is crucial (e.g., Baldwin & Fellingham, 2013; Browne & Draper, 2006; Gelman, 2006; Natarajan & Kass, 2000; Zitzmann, Lüdtke, & Robitzsch, 2015).”
BackgroundMethodInfluential
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.