Distinguishing between longitudinal dependence due to the effects of previous responses on subsequent responses and dependence due to unobserved heterogeneity is important in many disciplinesFor example, wheezing is an inflammatory reaction that may 'remodel' a child's airway structure and thereby affect the probability of future wheezing (state dependence)Alternatively, children could vary in their susceptibilities because of unobserved covariates such as genes (unobserved heterogeneity)For binary responses, distinguishing between state dependence and unobserved heterogeneity is typically accomplished by using dynamic/transition models that include both a lagged response and a random interceptNaive maximum likelihood estimators can be severely inconsistent because of two kinds of endogeneity problem: lack of independence of the initial response and the random intercept (the initial conditions problem) and lack of independence of the covariates and the random intercept (the endogenous covariates problem)We clarify and unify previous work on handling these problems in the disconnected literatures of statistics and econometrics, suggest improved methods, investigate the asymptotic performance of competing methods and provide practical recommendationsThe recommended methods are applied to longitudinal data on children's wheezing, where we investigate the extent of state dependence and unobserved heterogeneity and whether there is an effect of maternal smoking© 2013 Royal Statistical Society.

The classical missing at random (MAR) assumption, as defined by Rubin (Biometrika 63:581-592, 1976), is often not required for valid inference ignoring the missingness process. Neither are other assumptions sometimes believed to be necessary that result from misunderstandings of MAR. We discuss three strategies that allow us to use standard estimators (i.e., ignore missingness) in cases where missingness is usually considered to be non-ignorable: (1) conditioning on variables, (2) discarding more data, and (3) being protective of parameters.

Latent variable models are commonly used in medical statistics, although often not referred to under this name. In this paper we describe classical latent variable models such as factor analysis, item response theory, latent class models and structural equation models. Their usefulness in medical research is demonstrated using real data. Examples include measurement of forced expiratory flow, measurement of physical disability, diagnosis of myocardial infarction and modelling the determinants of clients' satisfaction with counsellors' interviews.

Composite links and exploded likelihoods are powerful yet simple tools for specifying a wide range of latent variable models. Applications considered include survival or duration models, models for rankings, small area estimation with census information, models for ordinal responses, item response models with guessing, randomized response models, unfolding models, latent class models with random effects, multilevel latent class models, models with log-normal latent variables, and zero-inflated Poisson models with random effects. Some of the ideas are illustrated by estimating an unfolding model for attitudes to female work participation. © 2007 The Psychometric Society.

We discuss prediction of random effects and of expected responses in multilevel generalized linear models. Prediction of random effects is useful for instance in small area estimation and disease mapping, effectiveness studies and model diagnostics. Prediction of expected responses is useful for planning, model interpretation and diagnostics. For prediction of random effects, we concentrate on empirical Bayes prediction and discuss three different kinds of standard errors; the posterior standard deviation and the marginal prediction error standard deviation (comparative standard errors) and the marginal sampling standard deviation (diagnostic standard error). Analytical expressions are available only for linear models and are provided in an appendix. For other multilevel generalized linear models we present approximations and suggest using parametric bootstrapping to obtain standard errors. We also discuss prediction of expectations of responses or probabilities for a new unit in a hypothetical cluster, or in a new (randomly sampled) cluster or in an existing cluster. The methods are implemented in gllamm and illustrated by applying them to survey data on reading proficiency of children nested in schools. Simulations are used to assess the performance of various predictions and associated standard errors for logistic random-intercept models under a range of conditions. © 2009 Royal Statistical Society.

In psychometrics, the canonical use of conditional likelihoods is for the Rasch model in measurement. Whilst not disputing the utility of conditional likelihoods in measurement, we examine a broader class of problems in psychometrics that can be addressed via conditional likelihoods. Specifically, we consider cluster-level endogeneity where the standard assumption that observed explanatory variables are independent from latent variables is violated. Here, "cluster" refers to the entity characterized by latent variables or random effects, such as individuals in measurement models or schools in multilevel models and "unit" refers to the elementary entity such as an item in measurement. Cluster-level endogeneity problems can arise in a number of settings, including unobserved confounding of causal effects, measurement error, retrospective sampling, informative cluster sizes, missing data, and heteroskedasticity. Severely inconsistent estimation can result if these challenges are ignored.

Wooldridge (2005) provided a simple and elegant solution to the initial conditions problem for dynamic nonlinear unobserved-effects models. His original auxiliary model includes the time-varying explanatory variables at each period. Unfortunately, a popular constrained version that includes within-means of the explanatory variables can be severely biased. We show that there are several ways to avoid this problem. © 2013 Elsevier B.V.

Latent variable modelling has gradually become an integral part of mainstream statistics and is currently used for a multitude of applications in different subject areas. Examples of 'traditional' latent variable models include latent class models, item-response models, common factor models, structural equation models, mixed or random effects models and covariate measurement error models. Although latent variables have widely different interpretations in different settings, the models have a very similar mathematical structure. This has been the impetus for the formulation of general modelling frameworks which accommodate a wide range of models. Recent developments include multilevel structural equation models with both continuous and discrete latent variables, multiprocess models and nonlinear latent variable models. © 2007 Board of the Foundation of the Scandinavian Journal of Statistics.

Gauss-Hermite quadrature is often used to evaluate and maximize the likelihood for random component probit models. Unfortunately, the estimates are biased for large cluster sizes and/or intraclass correlations. We show that adaptive quadrature largely overcomes these problems. We then extend the adaptive quadrature approach to general random coefficient models with limited and discrete dependent variables. The models can include several nested random effects (intercepts and coefficients) representing unobserved heterogeneity at different levels of a hierarchical dataset. The required multivariate integrals are evaluated efficiently using spherical quadrature rules. Simulations show that adaptive quadrature performs well in a wide range of situations. © 2004 Published by Elsevier B.V.

Generalized linear mixed models or multilevel regression models have become increasingly popular. Several methods have been proposed for estimating such models. However, to date there is no single method that can be assumed to work well in all circumstances in terms of both parameter recovery and computational efficiency. Stata's xt commands for two-level generalized linear mixed models (e.g., xtlogit) employ Gauss–Hermite quadrature to evaluate and maximize the marginal log likelihood. The method generally works very well, and often better than common contenders such as MQL and PQL, but there are cases where quadrature performs poorly. Adaptive quadrature has been suggested to overcome these problems in the two-level case. We have recently implemented a multilevel versionofthismethodin gllamm, a program that fits a large class of multilevel latent variable models including multilevel generalized linear mixed models. As far as we know, this is the first time that adaptive quadrature has been proposed for multilevel models. We show that adaptive quadrature works well in problems where ordinary quadrature fails. Furthermore, even when ordinary quadrature works, adaptive quadrature is often computationally more efficient since it requires fewer quadrature points to achieve the same precision.