Generalized linear mixed models: a review and some extensions

Breslow and Clayton (J Am Stat Assoc 88:9–25,1993) was, and still is, a highly influential paper mobilizing the use of generalized linear mixed models in epidemiology and a wide variety of fields. An important aspect is the feasibility in implementation through the ready availability of related software in SAS (SAS Institute, PROC GLIMMIX, SAS Institute Inc., URL , 2007), S-plus (Insightful Corporation, S-PLUS 8, Insightful Corporation, Seattle, WA, URL , 2007), and R (R Development Core Team, R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria, URL , 2006) for example, facilitating its broad usage. This paper reviews background to generalized linear mixed models and the inferential techniques which have been developed for them. To provide the reader with a flavor of the utility and wide applicability of this fundamental methodology we consider a few extensions including additive models, models for zero-heavy data, and models accommodating latent clusters.     

Modelling growth and decline in lung function in Duchenne's muscular dystrophy with an augmented linear mixed effects model

Summary.  Longitudinal modelling of lung function in Duchenne's muscular dystrophy is complicated by a mixture of both growth and decline in lung function within each subject, an unknown point of separation between these phases and significant heterogeneity between individual trajectories. Linear mixed effects models can be used, assuming a single changepoint for all cases; however, this assumption may be incorrect. The paper describes an extension of linear mixed effects modelling in which random changepoints are integrated into the model as parameters and estimated by using a stochastic EM algorithm. We find that use of this 'mixture modelling' approach improves the fit significantly.     

Three-step estimation in linear mixed models with skew-t distributions

Linear mixed models based on the normality assumption are widely used in health related studies. Although the normality assumption leads to simple, mathematically tractable, and powerful tests, violation of the assumption may easily invalidate the statistical inference. Transformation of variables is sometimes used to make normality approximately true. In this paper we consider another approach by replacing the normal distributions in linear mixed models by skew-t distributions, which account for skewness and heavy tails for both the random effects and the errors. The full likelihood-based estimator is often difficult to use, but a 3-step estimation procedure is proposed, followed by an application to the analysis of deglutition apnea duration in normal swallows. The example shows that skew-t models often entail more reliable inference than Gaussian models for the skewed data.     

Using a Box–Cox transformation in the analysis of longitudinal data with incomplete responses

We analyse longitudinal data on CD4 cell counts from patients who participated in clinical trials that compared two therapeutic treatments: zidovudine and didanosine. The investigators were interested in modelling the CD4 cell count as a function of treatment, age at base-line and disease stage at base-line. Serious concerns can be raised about the normality assumption of CD4 cell counts that is implicit in many methods and therefore an analysis may have to start with a transformation. Instead of assuming that we know the transformation (e.g. logarithmic) that makes the outcome normal and linearly related to the covariates, we estimate the transformation, by using maximum likelihood, within the Box–Cox family. There has been considerable work on the Box–Cox transformation for univariate regression models. Here, we discuss the Box–Cox transformation for longitudinal regression models when the outcome can be missing over time, and we also implement a maximization method for the likelihood, assumming that the missing data are missing at random.     

Modeling of covariance structures of random effects and random errors in linear mixed models

Abstract

In this paper, we discuss how to model the mean and covariancestructures in linear mixed models (LMMs) simultaneously. We propose a data-driven method to modelcovariance structures of the random effects and random errors in the LMMs. Parameter estimation in the mean and covariances is considered by using EM algorithm, and standard errors of the parameter estimates are calculated through Louis’ (1982 Louis, T.A. (1982). Finding observed information using the EM algorithm. J. Royal Stat. Soc. B 44:98–130. [Google Scholar]) information principle. Kenward’s (1987 Kenward, M.G. (1987). A method for comparing profiles of repeated measurements. Appl. Stat. 36:296–308.[Crossref], [Web of Science ®] , [Google Scholar]) cattle data sets are analyzed for illustration,and comparison to the literature work is made through simulation studies. Our numerical analysis confirms the superiority of the proposed method to existing approaches in terms of Akaike information criterion.     

A comparison of procedures to correct for base-line differences in the analysis of continuous longitudinal data: a case-study

Summary.  The main advantage of longitudinal studies is that they can distinguish changes over time within individuals (longitudinal effects) from differences between subjects at the start of the study (base-line characteristics; cross-sectional effects). Often, especially in observational studies, subjects are very heterogeneous at base-line, and one may want to correct for this, when doing inferences for the longitudinal trends. Three procedures for base-line correction are compared in the context of linear mixed models for continuous longitudinal data. All procedures are illustrated extensively by using data from an experiment which aimed at studying the relationship between the post-operative evolution of the functional status of elderly hip fracture patients and their preoperative neurocognitive status.     

Several nonparametric and semiparametric approaches to linear mixed model regression

Mixed models are powerful tools for the analysis of clustered data and many extensions of the classical linear mixed model with normally distributed response have been established. As with all parametric (P) models, correctness of the assumed model is critical for the validity of the ensuing inference. An incorrectly specified P means model may be improved by using a local, or nonparametric (NP), model. Two local models are proposed by a pointwise weighting of the marginal and conditional variance–covariance matrices. However, NP models tend to fit to irregularities in the data and may provide fits with high variance. Model robust regression techniques estimate mean response as a convex combination of a P and a NP model fit to the data. It is a semiparametric method by which incomplete or incorrectly specified P models can be improved by adding an appropriate amount of the NP fit. We compare the approximate integrated mean square error of the P, NP, and mixed model robust methods via a simulation study and apply these methods to two real data sets: the monthly wind speed data from countries in Ireland and the engine speed data.     

Computation aspects of the parameter estimates of linear mixed effects model in multivariate repeated measures set-up

The number of parameters mushrooms in a linear mixed effects (LME) model in the case of multivariate repeated measures data. Computation of these parameters is a real problem with the increase in the number of response variables or with the increase in the number of time points. The problem becomes more intricate and involved with the addition of additional random effects. A multivariate analysis is not possible in a small sample setting. We propose a method to estimate these many parameters in bits and pieces from baby models, by taking a subset of response variables at a time, and finally using these bits and pieces at the end to get the parameter estimates for the mother model, with all variables taken together. Applying this method one can calculate the fixed effects, the best linear unbiased predictions (BLUPs) for the random effects in the model, and also the BLUPs at each time of observation for each response variable, to monitor the effectiveness of the treatment for each subject. The proposed method is illustrated with an example of multiple response variables measured over multiple time points arising from a clinical trial in osteoporosis.     

Robust designs for linear mixed effects models

Summary.  In health sciences, medicine and social sciences linear mixed effects models are often used to analyse time-structured data. The search for optimal designs for these models is often hampered by two problems. The first problem is that these designs are only locally optimal. The second problem is that an optimal design for one model may not be optimal for other models. In this paper the maximin principle is adopted to handle both problems, simultaneously. The maximin criterion is formulated by means of a relative efficiency measure, which gives an indication of how much efficiency is lost when the uncertainty about the models over a prior domain of parameters is taken into account. The procedure is illustrated by means of three growth studies. Results are presented for a vocabulary growth study from education, a bone gain study from medical research and an epidemiological decline in height study. It is shown that, for the mixed effects polynomial models that are applied to these studies, the maximin designs remain highly efficient for different sets of models and combinations of parameter values.     

A Bayesian hierarchical model for categorical longitudinal data from a social survey of immigrants

Summary.  The paper investigates a Bayesian hierarchical model for the analysis of categorical longitudinal data from a large social survey of immigrants to Australia. Data for each subject are observed on three separate occasions, or waves, of the survey. One of the features of the data set is that observations for some variables are missing for at least one wave. A model for the employment status of immigrants is developed by introducing, at the first stage of a hierarchical model, a multinomial model for the response and then subsequent terms are introduced to explain wave and subject effects. To estimate the model, we use the Gibbs sampler, which allows missing data for both the response and the explanatory variables to be imputed at each iteration of the algorithm, given some appropriate prior distributions. After accounting for significant covariate effects in the model, results show that the relative probability of remaining unemployed diminished with time following arrival in Australia.     

Optimal design for linear regression models in the presence of heteroscedasticity caused by random coefficients

Random coefficients may result in heteroscedasticity of observations. For particular situations, where only one observation is available per individual, we derive optimal designs based on the geometry of the design locus.     

The Effect of Drop-Out on the Efficiency of Longitudinal Experiments

It is shown that drop-out often reduces the efficiency of longitudinal experiments considerably. In the framework of linear mixed models, a general, computationally simple method is provided, for designing longitudinal studies when drop-out is to be expected, such that there is little risk of large losses of efficiency due to the missing data. All the results are extensively illustrated using data from a randomized experiment with rats.     

A variance shift model for detection of outliers in the linear mixed measurement error models

ABSTRACT

In this paper, we extend a variance shift model, previously considered in the linear mixed models, to the linear mixed measurement error models using the corrected likelihood of Nakamura (1990 Nakamura, T. (1990). Corrected score function for errors in variables models: methodology and application to generalized linear models. Biometrika 77:127–137.[Crossref], [Web of Science ®] , [Google Scholar]). This model assumes that a single outlier arises from an observation with inflated variance. We derive the score test and the analogue of the likelihood ratio test, to assess whether the ith observation has inflated variance. A parametric bootstrap procedure is implemented to obtain empirical distributions of the test statistics. Finally, results of a simulation study and an example of real data are presented to illustrate the performance of proposed tests.     

Robust estimation in generalized linear mixed models

Generalized linear mixed models (GLMMs) are widely used to analyse non-normal response data with extra-variation, but non-robust estimators are still routinely used. We propose robust methods for maximum quasi-likelihood and residual maximum quasi-likelihood estimation to limit the influence of outlying observations in GLMMs. The estimation procedure parallels the development of robust estimation methods in linear mixed models, but with adjustments in the dependent variable and the variance component. The methods proposed are applied to three data sets and a comparison is made with the nonparametric maximum likelihood approach. When applied to a set of epileptic seizure data, the methods proposed have the desired effect of limiting the influence of outlying observations on the parameter estimates. Simulation shows that one of the residual maximum quasi-likelihood proposals has a smaller bias than those of the other estimation methods. We further discuss the equivalence of two GLMM formulations when the response variable follows an exponential family. Their extensions to robust GLMMs and their comparative advantages in modelling are described. Some possible modifications of the robust GLMM estimation methods are given to provide further flexibility for applying the method.     

Regression analysis of informative current status data with the semiparametric linear transformation model

Many methods have been developed in the literature for regression analysis of current status data with noninformative censoring and also some approaches have been proposed for semiparametric regression analysis of current status data with informative censoring. However, the existing approaches for the latter situation are mainly on specific models such as the proportional hazards model and the additive hazard model. Corresponding to this, in this paper, we consider a general class of semiparametric linear transformation models and develop a sieve maximum likelihood estimation approach for the inference. In the method, the copula model is employed to describe the informative censoring or relationship between the failure time of interest and the censoring time, and Bernstein polynomials are used to approximate the nonparametric functions involved. The asymptotic consistency and normality of the proposed estimators are established, and an extensive simulation study is conducted and indicates that the proposed approach works well for practical situations. In addition, an illustrative example is provided.     

A bayesian analysis of the mixed linear model

The mixed model is defined. The exact posterior distribution for the fixed effect vector is obtained. The exact posterior distribution for the error variance is obtained. The exact posterior mean and variance of a Bayesian estimator for the variances of random effects is also derived. All computations are non-iterative and avoid numerical integrations.     

Asymptotic bias in the linear mixed effects model under non-ignorable missing data mechanisms

Summary.  In longitudinal studies, missingness of data is often an unavoidable problem. Estimators from the linear mixed effects model assume that missing data are missing at random. However, estimators are biased when this assumption is not met. In the paper, theoretical results for the asymptotic bias are established under non-ignorable drop-out, drop-in and other missing data patterns. The asymptotic bias is large when the drop-out subjects have only one or no observation, especially for slope-related parameters of the linear mixed effects model. In the drop-in case, intercept-related parameter estimators show substantial asymptotic bias when subjects enter late in the study. Eight other missing data patterns are considered and these produce asymptotic biases of a variety of magnitudes.     

Screening for prostate cancer by using random-effects models

Summary. Random-effects models are used to screen male participants in a long-term longitudinal study for prostate cancer. By using posterior probabilities, each male can be classified into one of four diagnostic states for prostate disease: normal, benign prostatic hyperplasia, local cancer and metastatic cancer. Repeated measurements of prostate-specific antigen, collected when there was no clinical evidence of prostate disease, are used in the classification process. An individual's screening data are considered one repeated measurement at a time as his data are collected longitudinally over time. Posterior probabilities are calculated on the basis of data from other individuals with confirmed diagnoses of each of the four diagnostic states.