## Fitting Genetic Models Using Markov Chain Monte Carlo Algorithms with Bugs." Twin Research and Human Genetics 9 (2006)

Citations: | 4 - 2 self |

### BibTeX

@MISC{Bugs06fittinggenetic,

author = {Algorithms Bugs and Stéphanie M. Van Den Berg and Leo Beem and Dorret I. Boomsma},

title = {Fitting Genetic Models Using Markov Chain Monte Carlo Algorithms with Bugs." Twin Research and Human Genetics 9},

year = {2006}

}

### OpenURL

### Abstract

Maximum likelihood estimation techniques are widely used in twin and family studies, but soon reach computational boundaries when applied to highly complex models (e.g., models including geneby-environment interaction and gene–environment correlation, item response theory measurement models, repeated measures, longitudinal structures, extended pedigrees). Markov Chain Monte Carlo (MCMC) algorithms are very well suited to fit complex models with hierarchically structured data. This article introduces the key concepts of Bayesian inference and MCMC parameter estimation and provides a number of scripts describing relatively simple models to be estimated by the freely obtainable BUGS software. In addition, inference using BUGS is illustrated using a

### Citations

3918 |
Stochastic relaxation, Gibbs distribution and the Bayesian restoration of images
- Geman, Geman
- 1984
(Show Context)
Citation Context ... the integrals. The main ingredients of these methods are the Metropolis-Hastings algorithm and in particular its special case, the Gibbs sampler. The Gibbs sampler has its roots in image processing (=-=Geman & Geman, 1984-=-). The algorithms found their application especially in Bayesian models, although they have also been used in classical frequentist models. From a purely formal point of view, the main difference is t... |

814 |
Sampling-Based Approaches to Calculating Marginal Densities
- Gelfand, Smith
- 1990
(Show Context)
Citation Context ...by itself is not easy to sample from directly. It may be possible, however, to sample directly from the conditional distributions P(η | ξ,Y) and P(ξ | η, Y). This approach is known as Gibbs sampling (=-=Gelfand & Smith, 1990-=-; Geman & Geman, 1984), which is a special case of the general class of MCMC algorithms. When, however, the conditional posterior density of a subset of the parameters is not easy or even impossible t... |

477 | Bayesian Inference in Statistical Analysis - Box, Tiao - 1973 |

163 | Bayesian measures of model complexity and fit - Spiegelhalter, Best, et al. |

150 |
Multilevel and Longitudinal Modeling Using Stata
- Rabe-Hesketh, Skrondal
- 2005
(Show Context)
Citation Context ...te data such as dichotomous or ordered categorical data are often fitted by threshold models. Packages that allow the estimation of generalized linear mixed models (e.g., the GLLAMM routine in STATA, =-=Rabe-Hesketh & Skrondal, 2005-=-) and most SEM packages can handle threshold models or other models where latent variables are assumed underlying observed dichotomous Twin Research and Human Genetics Volume 9 Number 3 pp. 334–342 or... |

101 | Generalized latent variable modeling: Multilevel, longitudinal and structural equation models. Boca Raton, FL - Skrondal, Rabe-Hesketh - 2004 |

81 | Netherlands Twin Register: A focus on longitudinal research - Boomsma, Vink, et al. - 2002 |

58 |
Generalized, Linear and Mixed Models
- McCulloch, Searle
- 2001
(Show Context)
Citation Context ...e methods even for nonnormal continuous variables. A natural extension to discrete data are generalized linear mixed models (GLMM), which are generalized linear models with continuous random effects (=-=McCulloch & Searle, 2001-=-), or generalized linear latent variable models (Skrondal & RabeHesketh, 2004). In genetics, discrete data such as dichotomous or ordered categorical data are often fitted by threshold models. Package... |

31 |
The inheritance of liability to certain diseases, estimated from the incidence among relatives
- Falconer
- 1965
(Show Context)
Citation Context ...eral approach to model reparameterization to reduce the correlations among parameters in the posterior distribution is presented in Appendix A. Threshold Models The threshold model (Crittenden, 1961; =-=Falconer, 1965-=-) is often applied in order to estimate the heritability of continuous but categorically measured traits. In this model, an underlying continuous trait is defined, usually termed liability, with a thr... |

26 |
Comparison of the biometrical genetical, MAVA, and classical approaches to the analysis of human behavior. Psychol Bull
- Jinks, Fulker
- 1970
(Show Context)
Citation Context ...ymbol and subscripts are identical. The additive genetic effect is split into a1 i and a2 ij’ in order to model the different genetic correlations among monozygotic (MZ) and dizygotic (DZ) twins (cf. =-=Jinks & Fulker, 1970-=-). The genetic variance is the same for DZ and MZ twins, while the genetic covariance in MZ twins is twice as large as in DZ twins. We now incorporate the restriction VAR(a1) = VAR(a2) = 1/2 σ2 a , so... |

25 |
Efficient parameterisations for normal linear mixed models
- GELFAND, SAHU, et al.
- 1995
(Show Context)
Citation Context ...centred representation. Such a formulation may lead to more efficient MCMC sampling, as it can reduce or sometimes effectively nullify the correlations among parameters in the posterior distribution (=-=Gelfand et al., 1995-=-). In mixed models the random effects are usually specified as having a zero mean (see above). In the centered specifications, the distribution of random effects may contain means that are not equal t... |

17 |
Using LISREL to analyze genetic and environmental covariance structure
- Boomsma, Molenaar
- 1986
(Show Context)
Citation Context ...006 Fitting Models Using MCMC Algorithms With BUGS cient to model the sums and differences for each twin pair i instead of modeling the observed phenotypic values (Robert & Casella, 2004, p. 396; cf. =-=Boomsma & Molenaar, 1987-=-): S i = Phenotype i1 + Phenotype i2 and D i = Phenotype i1 – Phenotype i 2 . The covariance of the transformed phenotypes is cov(P1 + P2, P1 – P2) = var (P1) – cov(P1, P2) + cov(P1, P2) – var(P2) = 0... |

11 | Markov chain Monte Carlo approaches to analysis of genetic and environmental components of human developmental change and
- Eaves, Erkanli
- 2003
(Show Context)
Citation Context ...now often called Markov Chain Monte Carlo (MCMC) methods or algorithms. The MCMC methodology has been successively applied for twin models with gene–environment (G–E) correlation and G–E interaction (=-=Eaves & Erkanli, 2003-=-), a multivariate threshold model for polytomous data (van den Berg et al., 2006) and twin models that include measurement models based on item response theory (IRT; Eaves et al., 2005, 2004; van den ... |

7 |
Bayesian data analysis (2nd
- Gelman, Carlin, et al.
- 2003
(Show Context)
Citation Context ...t the estimates of interest. Therefore we extend the model by adding relatively noninformative uniform distributions for the square-root of the variance components (the standard deviations, following =-=Gelman et al., 2004-=-), σ e ~ U(0, 100), σ c ~ U(0, 100), σ a ~ U(0, 100), and a locally ‘flat’ normal prior for the population mean, µ ~ N(0, 100), so that the joint posterior distribution is mainly determined by the lik... |

5 |
LB: An interpretation of familial aggregation based on multiple genetic and environmental factors
- Crittenden
- 1961
(Show Context)
Citation Context ...σ2 e ). A more general approach to model reparameterization to reduce the correlations among parameters in the posterior distribution is presented in Appendix A. Threshold Models The threshold model (=-=Crittenden, 1961-=-; Falconer, 1965) is often applied in order to estimate the heritability of continuous but categorically measured traits. In this model, an underlying continuous trait is defined, usually termed liabi... |

4 |
Covariance components models for longitudinal family data
- Burton, Scurrah, et al.
- 2005
(Show Context)
Citation Context ... swiftly, particularly models with highly complex dependency structures, such as latent growth models for repeated measures on traits that are measured using a limited set of items (see, for example, =-=Burton et al., 2005-=-; van den Berg et al., submitted). Future work should, however, focus on alternative parameterizations for specific types of models to speed up MCMC estimation and on the assessment of model fit. Ackn... |

4 |
Application of Bayesian inference using Gibbs sampling to item-response theory modeling of multi-symptom genetic data. Behavior Genetics 2005
- Eaves, Erkanli, et al.
(Show Context)
Citation Context ...interaction (Eaves & Erkanli, 2003), a multivariate threshold model for polytomous data (van den Berg et al., 2006) and twin models that include measurement models based on item response theory (IRT; =-=Eaves et al., 2005-=-, 2004; van den Berg et al., submitted). MCMC estimation can be easily performed using the BUGS family of software programs that is currently freely obtainable Received 11 November, 2005; accepted 17 ... |

4 | Genetic and environmental influences on the relative timing of pubertal change - Eaves, Silberg, et al. - 2004 |