Youngjo Lee

Conditional and Marginal Models: Another View

There has existed controversy about the use of marginal and conditional models, particularly in the analysis of data from longitudinal studies. We show that alleged differences in the behavior of parameters in so-called marginal and conditional models are based on a failure to compare like with like. In particular, these seemingly apparent differences are meaningless because they are mainly caused by preimposed unidentifiable constraints on the random effects in models. We discuss the advantages of conditional models over marginal models. We regard the conditional model as fundamental, from which marginal predictions can be made.

Robust design via generalized linear models

A single data transformation may fail to satisfy all the required properties necessary for an analysis. With generalized linear models (GLMs), the identification of the mean-variance relationship and the choice of the scale on which the effects are to be measured can be done separately, overcoming the shortcomings of the data-transformation approach. GLMs also provide an extension of the response surface approach. In this paper, we set out the current status of the GLM approach to the analysis of data from quality-improvement experiments and discuss its merits.

On weighting of bivariate margins in pairwise likelihood

Composite and pairwise likelihood methods have recently been increasingly used. For clustered data with varying cluster sizes, we study asymptotic relative efficiencies for various weighted pairwise likelihoods, with weight being a function of cluster size. For longitudinal data, we also study weighted pairwise likelihoods with weights that can depend on lag. Good choice of weights are needed to avoid the undesirable behavior of estimators with low efficiency. Some analytic results are obtained using the multivariate normal distribution. For clustered data, a practically good choice of weight is obtained after study of relative efficiencies for an exchangeable multivariate normal model; they are different from weights that had previously been suggested. For longitudinal data, there are advantages to only include bivariate margins of adjacent or nearly adjacent pairs in the weighted pairwise likelihood.

Estimating Frailty Models via Poisson Hierarchical Generalized Linear Models

Frailty models extend proportional hazards models to multivariate survival data. Hierarchical-likelihood provides a simple unified framework for various random effect models such as hierarchical generalized linear models, frailty models, and mixed linear models with censoring. Wereview the hierarchical-likelihood estimation methods for frailty models. Hierarchical-likelihood for frailty models can be expressed as that for Poisson hierarchical generalized linear models. Frailty models can thus be fitted using Poisson hierarchical generalized linear models. Properties of the new methodology are demonstrated by simulation. The new method reduces the bias of maximum likelihood and penalized likelihood estimates.

Modelling and analysing correlated non-normal data:

We introduce a model class that includes many types of correlation structures for non-Gaussian models. We then show how to check the underlying model assumptions to discriminate between different correlation patterns and demonstrate how to select suitable models. Strawberry data are used to discuss the choice between fixed- and random-effect models for the fertility effect in agricultural experiments. Prostate-cancer data are used to demonstrate the method applied to the analysis of longitudinal studies and Scottish lip-cancer data to illustrate an application to spatial statistics.

Transcranial Direct Current Stimulation to Lessen Neuropathic Pain After Spinal Cord Injury A Mechanistic PET Study

Background. It is suggested that transcranial direct current stimulation (tDCS) can produce lasting changes in corticospinal excitability and can potentially be used for the treatment of neuropathic pain. However, the detailed mechanisms underlying the effects of tDCS are unknown. Objective. We investigated the underlying neural mechanisms of tDCS for chronic pain relief using [18F]-fluorodeoxyglucose positron emission tomography ([18F]FDG-PET). Methods. Sixteen patients with neuropathic pain (mean age 44.1 ± 8.6 years, 4 females) due to traumatic spinal cord injury received sham or active anodal stimulation of the motor cortex using tDCS for 10 days (20 minutes, 2 mA, twice a day). The effect of tDCS on regional cerebral glucose metabolism was evaluated by [18F]FDG-PET before and after tDCS sessions. Results. There was a significant decrease in the numeric rating scale scores for pain, from 7.6 ± 0.5 at baseline to 5.9 ± 1.8 after active tDCS (P = .016). We found increased metabolism in the medulla and decreased metabolism in the left dorsolateral prefrontal cortex after active tDCS treatment compared with the changes induced by sham tDCS. Additionally, an increase in metabolism after active tDCS was observed in the subgenual anterior cingulate cortex and insula. Conclusion. The results of this study suggest that anodal stimulation of the motor cortex using tDCS can modulate emotional and cognitive components of pain and normalize excessive attention to pain and pain-related information.

Two ways of modelling overdispersion in non‐normal data

For non-normal data assumed to have distributions, such as the Poisson distribution, which have an a priori dispersion parameter, there are two ways of modelling overdispersion: by a quasi-likelihood approach or with a random-effect model. The two approaches yield different variance functions for the response, which may be distinguishable if adequate data are available. The epilepsy data of Thall and Vail and the fabric data of Bissell are used to exemplify the ideas.

Cortical and white matter alterations in patients with neuropathic pain after spinal cord injury.

Neuropathic pain is one of the major problems of patients with spinal cord injury (SCI), which remains refractory to treatment despite a variety of therapeutic approach. Multimodal neuroimaging could provide complementary information for brain mechanisms underlying neuropathic pain, which could be based on development of more effective treatment strategies. Ten patients suffering from chronic neuropathic pain after SCI and 10 healthy controls underwent FDG-PET, T1-anatomical MRI and diffusion tensor imaging. We found decreases of both metabolism and the gray matter volume in the left dorsolateral prefrontal cortex in patients compared to healthy controls, as well as hypometabolism in the medial prefrontal cortex and gray matter volume loss in bilateral anterior insulae and subgenual anterior cingulate cortices. These brain regions are generally known to participate in pain modulation by affective and cognitive processes. Decreases of mean diffusivity (MD) in the right internal capsule including, cerebral peduncle, pre-and post-central white matter, and prefrontal white matter as components of the corticospinal and thalamocortical tracts were demonstrated in patients. Further, lower MD value of prefrontal white matter was correlated with decreased metabolism of medial prefrontal cortex in patients. These results indicated that white matter changes imply abnormal pain modulation in patients as well as motor impairment. Our study showed the functional and structural multimodal imaging modality commonly identified the possible abnormalities in the brain regions participating pain modulation in neuropathic pain. Multifaceted imaging studies in neuropathic pain could be useful elucidating precise mechanisms of persistent pain, and providing future directions for treatment.

Super-sparse principal component analyses for high-throughput genomic data

BackgroundPrincipal component analysis (PCA) has gained popularity as a method for the analysis of high-dimensional genomic data. However, it is often difficult to interpret the results because the principal components are linear combinations of all variables, and the coefficients (loadings) are typically nonzero. These nonzero values also reflect poor estimation of the true vector loadings; for example, for gene expression data, biologically we expect only a portion of the genes to be expressed in any tissue, and an even smaller fraction to be involved in a particular process. Sparse PCA methods have recently been introduced for reducing the number of nonzero coefficients, but these existing methods are not satisfactory for high-dimensional data applications because they still give too many nonzero coefficients.ResultsHere we propose a new PCA method that uses two innovations to produce an extremely sparse loading vector: (i) a random-effect model on the loadings that leads to an unbounded penalty at the origin and (ii) shrinkage of the singular values obtained from the singular value decomposition of the data matrix. We develop a stable computing algorithm by modifying nonlinear iterative partial least square (NIPALS) algorithm, and illustrate the method with an analysis of the NCI cancer dataset that contains 21,225 genes.ConclusionsThe new method has better performance than several existing methods, particularly in the estimation of the loading vectors.

Estimation of breeding values for mean and dispersion, their variance and correlation using double hierarchical generalized linear models

The possibility of breeding for uniform individuals by selecting animals expressing a small response to environment has been studied extensively in animal breeding. Bayesian methods for fitting models with genetic components in the residual variance have been developed for this purpose, but have limitations due to the computational demands. We use the hierarchical (h)-likelihood from the theory of double hierarchical generalized linear models (DHGLM) to derive an estimation algorithm that is computationally feasible for large datasets. Random effects for both the mean and residual variance parts of the model are estimated together with their variance/covariance components. An important feature of the algorithm is that it can fit a correlation between the random effects for mean and variance. An h-likelihood estimator is implemented in the R software and an iterative reweighted least square (IRWLS) approximation of the h-likelihood is implemented using ASReml. The difference in variance component estimates between the two implementations is investigated, as well as the potential bias of the methods, using simulations. IRWLS gives the same results as h-likelihood in simple cases with no severe indication of bias. For more complex cases, only IRWLS could be used, and bias did appear. The IRWLS is applied on the pig litter size data previously analysed by Sorensen & Waagepetersen (2003) using Bayesian methodology. The estimates we obtained by using IRWLS are similar to theirs, with the estimated correlation between the random genetic effects being -0·52 for IRWLS and -0·62 in Sorensen & Waagepetersen (2003).