Multivariate methods for clustered binary data with multiple subclasses, with application to binary longitudinal data.

Clustered binary data occur frequently in biostatistical work. Several approaches have been proposed for the analysis of clustered binary data. In Rosner (1984, Biometrics 40, 1025-1035), a polychotomous logistic regression model was proposed that is a generalization of the beta-binomial distribution and allows for unit- and subunit-specific covariates, while controlling for clustering effects. One assumption of this model is that all pairs of subunits within a cluster are equally correlated. This is appropriate for ophthalmologic work where clusters are generally of size 2, but may be inappropriate for larger cluster sizes. A beta-binomial mixture model is introduced to allow for multiple subclasses within a cluster and to estimate odds ratios relating outcomes for pairs of subunits within a subclass as well as in different subclasses. To include covariates, an extension of the polychotomous logistic regression model is proposed, which allows one to estimate effects of unit-, class-, and subunit-specific covariates, while controlling for clustering using the beta-binomial mixture model. This model is applied to the analysis of respiratory symptom data in children collected over a 14-year period in East Boston, Massachusetts, in relation to maternal and child smoking, where the unit is the child and symptom history is divided into early-adolescent and late-adolescent symptom experience.     

Nonparametric variance estimation in the analysis of microarray data: a measurement error approach

We investigate the effects of measurement error on the estimationof nonparametric variance functions. We show that either ignoringmeasurement error or direct application of the simulation extrapolation,SIMEX, method leads to inconsistent estimators. Nevertheless,the direct SIMEX method can reduce bias relative to a naiveestimator. We further propose a permutation SIMEX method thatleads to consistent estimators in theory. The performance ofboth the SIMEX methods depends on approximations to the exactextrapolants. Simulations show that both the SIMEX methods performbetter than ignoring measurement error. The methodology is illustratedusing microarray data from colon cancer patients.     

DiscoverySpace: an interactive data analysis application

DiscoverySpace is a graphical application for bioinformatics data analysis. Users can seamlessly traverse references between biological databases and draw together annotations in an intuitive tabular interface. Datasets can be compared using a suite of novel tools to aid in the identification of significant patterns. DiscoverySpace is of broad utility and its particular strength is in the analysis of serial analysis of gene expression (SAGE) data. The application is freely available online.     

Identifying multiple changepoints in heterogeneous binary data with an application to molecular genetics

Identifying changepoints is an important problem in molecular genetics. Our motivating example is from cancer genetics where interest focuses on identifying areas of a chromosome with an increased likelihood of a tumor suppressor gene. Loss of heterozygosity (LOH) is a binary measure of allelic loss in which abrupt changes in LOH frequency along the chromosome may identify boundaries indicative of a region containing a tumor suppressor gene. Our interest was on testing for the presence of multiple changepoints in order to identify regions of increased LOH frequency. A complicating factor is the substantial heterogeneity in LOH frequency across patients, where some patients have a very high LOH frequency while others have a low frequency. We develop a procedure for identifying multiple changepoints in heterogeneous binary data. We propose both approximate and full maximum-likelihood approaches and compare these two approaches with a naive approach in which we ignore the heterogeneity in the binary data. The methodology is used to estimate the pattern in LOH frequency on chromosome 13 in esophageal cancer patients and to isolate an area of inflated LOH frequency on chromosome 13 which may contain a tumor suppressor gene. Using simulations, we show that our approach works well and that it is robust to departures from some key modeling assumptions.     

New approach to the statistical analysis of cardiovascular data.

Fourier-based approaches to analysis of variability of R-R intervals or blood pressure typically compute power in a given frequency band (e.g., 0.01-0.07 Hz) by aggregating the power at each constituent frequency within that band. This paper describes a new approach to the analysis of these data. We propose to partition the blood pressure variability spectrum into more narrow components by computing power in 0.01-Hz-wide bands. Therefore, instead of a single measure of variability in a specific frequency interval, we obtain several measurements. The approach generates a more complex data structure that requires a careful account of the nested repeated measures. We briefly describe a statistical methodology based on generalized estimating equations that suitably handles this more complex data structure. To illustrate the methods, we consider systolic blood pressure data collected during psychological and orthostatic challenge. We compare the results with those obtained using the conventional methods to compute blood pressure variability, and we show that our approach yields more efficient results and more powerful statistical tests. We conclude that this approach may allow a more thorough analysis of cardiovascular parameters that are measured under different experimental conditions, such as blood pressure or heart rate variability.     

Human error and the problem of causality in analysis of accidents

Present technology is characterized by complexity, rapid change and growing size of technical systems. This has caused increasing concern with the human involvement in system safety. Analyses of the major accidents during recent decades have concluded that human errors on part of operators, designers or managers have played a major role. There are, however, several basic problems in analysis of accidents and identification of human error. This paper addresses the nature of causal explanations and the ambiguity of the rules applied for identification of the events to include in analysis and for termination of the search for 'causes'. In addition, the concept of human error is analysed and its intimate relation with human adaptation and learning is discussed. It is concluded that identification of errors as a separate class of behaviour is becoming increasingly difficult in modern work environments. The influence of this change on the control of safety of large-scale industrial systems is discussed.     

The geneis and mechanism of human error in accidents resulting latent danger.

Six accidental cases caused by overlooking latent environmental danger were analyzed and the geneses and mechanism of the human error involved were discussed. In all cases detection of the danger was easy or indications by some warning were distinct. In a typical case, a lineman, after cleaning the insulators of a suspended line, proceeded toward the danger zone of another line charged with current on reaching the ground. He disregarded the warning flag, and received a shock and fell to the ground. It was also common in all the other cases that victims were conducting relatively easy or habitual activities and responded, being almost unaware of the unusual physical characteristics of the environment, to its more apparent aspects. These errors were relevant to experimental results by the author that subjects had great difficulty in identifying the rule of presentation of signal figures if the key item and the rule, either serial or positional, were different from foregoing trials. It is suggested that an individual fails to see an external object that has little function value in regard to the content of the current behavior under influence of personal habitual experiences and the theories of human error need to be reconstruct as the basis of ecological knowledge about them.     

Cost-efficient study designs for binary response data with Gaussian covariate measurement error.

When mismeasurement of the exposure variable is anticipated, epidemiologic cohort studies may be augmented to include a validation study, where a small sample of data relating the imperfect exposure measurement method to the better method is collected. Optimal study designs (i.e., least expensive subject to specified power constraints) are developed that give the overall sample size and proportion of the overall sample size allocated to the validation study. If better exposure measurements can be collected on a sample of subjects, an optimal design can be suggested that conforms to realistic budgetary constraints. The properties of three designs--those that include an internal validation study, those where the validated subsample is derived from subjects external to the primary investigation, and those that use the better method of exposure assessment on all subjects--are compared. The proportion of overall study resources allocated to the validation substudy increases with increasing sample disease frequency, decreasing unit cost of the superior exposure measurement relative to the imperfect one, increasing unit cost of outcome ascertainment, increasing distance between two alternative values of the relative risk between which the study is designed to discriminate, and increasing magnitude of hypothesized values. This proportion also depends in a nonlinear fashion on the severity of measurement error, and when the validation study is internal, measurement error reaches a point after which the optimal design is the smaller, fully validated one.     

Zygosity diagnosis in the absence of genotypic data: an approach using latent class analysis.

For zygosity diagnosis in the absence of genotypic data, or in the recruitment phase of a twin study where only single twins from same-sex pairs are being screened, or to provide a test for sample duplication leading to the false identification of a dizygotic pair as monozygotic, the appropriate analysis of respondents' answers to questions about zygosity is critical. Using data from a young adult Australian twin cohort (N = 2094 complete pairs and 519 singleton twins from same-sex pairs with complete responses to all zygosity items), we show that application of latent class analysis (LCA), fitting a 2-class model, yields results that show good concordance with traditional methods of zygosity diagnosis, but with certain important advantages. These include the ability, in many cases, to assign zygosity with specified probability on the basis of responses of a single informant (advantageous when one zygosity type is being oversampled); and the ability to quantify the probability of misassignment of zygosity, allowing prioritization of cases for genotyping as well as identification of cases of probable laboratory error. Out of 242 twins (from 121 like-sex pairs) where genotypic data were available for zygosity confirmation, only a single case was identified of incorrect zygosity assignment by the latent class algorithm. Zygosity assignment for that single case was identified by the LCA as uncertain (probability of being a monozygotic twin only 76%), and the co-twin's responses clearly identified the pair as dizygotic (probability of being dizygotic 100%). In the absence of genotypic data, or as a safeguard against sample duplication, application of LCA for zygosity assignment or confirmation is strongly recommended.     

A simple method for the analysis of clustered binary data.

A simple method for comparing independent groups of clustered binary data with group-specific covariates is proposed. It is based on the concepts of design effect and effective sample size widely used in sample surveys, and assumes no specific models for the intracluster correlations. It can be implemented using any standard computer program for the analysis of independent binary data after a small amount of preprocessing. The method is applied to a variety of problems involving clustered binary data: testing homogeneity of proportions, estimating dose-response models and testing for trend in proportions, and performing the Mantel-Haenszel chi-squared test for independence in a series of 2 x 2 tables and estimating the common odds ratio and its variance. Illustrative applications of the method are also presented.     

Intrinsic mode entropy based on multivariate empirical mode decomposition and its application to neural data analysis

Entropy, a measure of the regularity of a time series, has long been used to quantify the complexity of brain dynamics. Given the multiple spatiotemporal scales inherent in the brain, traditional entropy analysis based on a single scale is not adequate to accurately describe the underlying nonlinear dynamics. Intrinsic mode entropy (IMEn) is a recent development with appealing properties to estimate entropy over multiple time scales. It is a multiscale entropy measure that computes sample entropy (SampEn) over different scales of intrinsic mode functions extracted by empirical mode decomposition (EMD) method. However, it suffers from both mode-misalignment and mode-mixing problems when applied to multivariate time series data. In this paper, we address these two problems by employing the recently introduced multivariate empirical mode decomposition (MEMD). First, we extend the MEMD to multi-channel multi-trial neural data to ensure the IMEn matched at different scales. Second, for the discriminant analysis of IMEn, we propose to improve the discriminative ability by including variance that has not been used before in entropy analysis. Finally, we apply the proposed approach to the multi-electrode local field potentials (LFPs) simultaneously collected from visual cortical areas of macaque monkeys while performing a generalized flash suppression task. The results have shown that the entropy of LFP is indeed scale-dependent and is closely related to the perceptual conditions. The discriminative results of the perceptual conditions, revealed by support vector machine, show that the accuracy based on IMEn and variance reaches 83.05%, higher than that only by IMEn (76.27%). These results suggest that our approach is sensitive to capture the complex dynamics of neural data.     

An autologistic model for the genetic analysis of familial binary data.

Regressive logistic models specify the probability distribution of familial binary traits by conditioning each individual's phenotype on those of preceding relatives; therefore, the expression of the joint probability of the familial data necessitates ordering the observations. In the present paper, we propose an autologistic model of this familial dependence structure, which does not require specification of a particular ordering of the phenotypic observations. Genetic effects are introduced into the model in order to perform segregation analysis that is aimed at detecting the role of a major locus in the expression of familial phenotypes. In this model, the conditional probabilities have a logistic form, and large patterns of dependence between relatives can be considered with a simple interpretation of the parameters measuring the relationship between two phenotypes. The model is compared with the regressive logistic approach in terms of odds ratios and by using a simulation study.     

MUNIN: Application of three-way decomposition to the analysis of heteronuclear NMR relaxation data**

MUNIN (Multidimensional NMR Spectra Interpretation), a recently introduced approach exploiting the mathematical concept of three-way decomposition, is proposed for separation and quantitative relaxation measurements of strongly overlapped resonances in sets of heteronuclear two-dimensional spectra that result from typical relaxation experiments. The approach is general and may also be applied to sets of two-dimensional spectra with arbitrary modulation along the third dimension (e.g., J-coupling, diffusion). Here, the method is applied for the analysis of ¹⁵N rotating frame relaxation data.     

Studies on rationales for an expert system approach to the interpretation of protein sequence data. Preliminary analysis of the human epidermal growth factor receptor

The molecular modelling of larger proteins benefits from a preliminary analysis of the sequence to identify regions of potential structural and functional importance. In this study the sequence of the epidermal growth factor receptor has been analysed using a variety of established methods and novel procedures developed for the study of weak internal and external homologies and for the use of homologous sequences in the prediction of secondary and super-secondary structures. The procedures explored here are potentially suitable for incorporation into an expert system for the initial investigation of protein sequence data.     

Differential coexpression analysis using microarray data and its application to human cancer

MOTIVATION: Microarrays have been used to identify differential expression of individual genes or cluster genes that are coexpressed over various conditions. However, alteration in coexpression relationships has not been studied. Here we introduce a model for finding differential coexpression from microarrays and test its biological validity with respect to cancer. RESULTS: We collected 10 published gene expression datasets from cancers of 13 different tissues and constructed 2 distinct coexpression networks: a tumor network and normal network. Comparison of the two networks showed that cancer affected many coexpression relationships. Functional changes such as alteration in energy metabolism, promotion of cell growth and enhanced immune activity were accompanied with coexpression changes. Coregulation of collagen genes that may control invasion and metastatic spread of tumor cells was also found. Cluster analysis in the tumor network identified groups of highly interconnected genes related to ribosomal protein synthesis, the cell cycle and antigen presentation. Metallothionein expression was also found to be clustered, which may play a role in apoptosis control in tumor cells. Our results show that this model would serve as a novel method for analyzing microarrays beyond the specific implications for cancer.