Flexible parametric models for random‐effects distributions

It is commonly assumed that random effects in hierarchical models follow a normal distribution. This can be extremely restrictive in practice. We explore the use of more flexible alternatives for this assumption, namely the t distribution, and skew extensions to the normal and t distributions, implemented using Markov Chain Monte Carlo methods. Models are compared in terms of parameter estimates, deviance information criteria, and predictive distributions. These methods are applied to examples in meta-analysis and health-professional variation, where the distribution of the random effects is of direct interest. The results highlight the importance of allowing for potential skewing and heavy tails in random-effects distributions, especially when estimating a predictive distribution. We describe the extension of these random-effects models to the bivariate case, with application to a meta-analysis examining the relationship between treatment effect and baseline response. We conclude that inferences regarding the random effects can crucially depend on the assumptions made and recommend using a distribution, such as those suggested here, which is more flexible than the normal.

[1]  N. Laird,et al.  Meta-analysis in clinical trials. , 1986, Controlled clinical trials.

[2]  S. Thompson,et al.  Clustering by health professional in individually randomised trials , 2005, BMJ : British Medical Journal.

[3]  F. Vaida,et al.  Proportional hazards model with random effects. , 2000, Statistics in medicine.

[4]  M. Aitkin,et al.  Meta-analysis by random effect modelling in generalized linear models. , 1999, Statistics in medicine.

[5]  H Goldstein,et al.  A multilevel model framework for meta-analysis of clinical trials with binary outcomes. , 2000, Statistics in medicine.

[6]  A. Sheiham,et al.  Fluoride toothpastes for preventing dental caries in children and adolescents. , 2003, The Cochrane database of systematic reviews.

[7]  T Stijnen,et al.  Baseline risk as predictor of treatment benefit: three clinical meta-re-analyses. , 2000, Statistics in medicine.

[8]  Simon G Thompson,et al.  The use of random effects models to allow for clustering in individually randomized trials , 2005, Clinical trials.

[9]  A. Gelman.  Prior distributions for variance parameters in hierarchical models (comment on article by Browne and Draper) , 2004 .

[10]  A Whitehead,et al.  Meta‐analysis of ordinal outcomes using individual patient data , 2001, Statistics in medicine.

[11]  Y. Ohashi,et al.  A Bayesian hierarchical survival model for the institutional effects in a multi-centre cancer clinical trial. , 1998, Statistics in medicine.

[12]  D J Spiegelhalter,et al.  Flexible random‐effects models using Bayesian semi‐parametric models: applications to institutional comparisons , 2007, Statistics in medicine.

[13]  D. Mehrotra.  Non-iterative robust estimators of variance components in within-subject designs. , 1997, Statistics in medicine.

[14]  C S Berkey,et al.  Multiple-outcome meta-analysis of clinical trials. , 1996, Statistics in medicine.

[15]  S G Thompson,et al.  Analysing the relationship between treatment effect and underlying risk in meta-analysis: comparison and development of approaches. , 2000, Statistics in medicine.

[16]  M. C. Jones,et al.  A skew extension of the t‐distribution, with applications , 2003 .

[17]  Simon G Thompson,et al.  Modelling Multivariate Outcomes in Hierarchical Data, with Application to Cluster Randomised Trials , 2006, Biometrical journal. Biometrische Zeitschrift.

[18]  R J Carroll,et al.  Flexible Parametric Measurement Error Models , 1999, Biometrics.

[19]  J. Kalbfleisch,et al.  The effects of mixture distribution misspecification when fitting mixed-effects logistic models , 1992 .

[20]  H. Goldstein.  Multilevel Statistical Models , 2006 .

[21]  R Z Omar,et al.  Bayesian methods of analysis for cluster randomized trials with binary outcome data. , 2001, Statistics in medicine.

[22]  M. Steel,et al.  On Bayesian Modelling of Fat Tails and Skewness , 1998 .

[23]  H C Van Houwelingen,et al.  A bivariate approach to meta-analysis. , 1993, Statistics in medicine.

[24]  P Jacklin,et al.  Joint teleconsultations (virtual outreach) versus standard outpatient appointments for patients referred by their general practitioner for a specialist opinion: a randomised trial , 2002, The Lancet.

[25]  Marie Davidian,et al.  Consequences of misspecifying assumptions in nonlinear mixed effects models , 2000 .

[26]  Douglas G Altman,et al.  The relation between treatment benefit and underlying risk in meta-analysis , 1996, BMJ.

[27]  Brian Everitt,et al.  Statistical analysis of medical data : new developments , 2000 .

[28]  D. Buckle.  Bayesian Inference for Stable Distributions , 1995 .

[29]  Yasuo Ohashi,et al.  Overlap coefficient for assessing the similarity of pharmacokinetic data between ethnically different populations , 2005, Clinical trials.

[30]  D J Spiegelhalter,et al.  Bayesian approaches to random-effects meta-analysis: a comparative study. , 1995, Statistics in medicine.

[31]  D. Burr,et al.  A Bayesian Semiparametric Model for Random-Effects Meta-Analysis , 2005 .

[32]  Hani Doss,et al.  A meta‐analysis of studies on the association of the platelet PlA polymorphism of glycoprotein IIIa and risk of coronary heart disease , 2003, Statistics in medicine.

[33]  A. Azzalini,et al.  Statistical applications of the multivariate skew normal distribution , 2009, 0911.2093.