ISSN:0162-1459    

DOI:10.1080/01621459.1996.10476659

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

Statisticians have contributed enormously to the conceptualization, development, and success of case-control methods for the study of disease causation and prevention. This article reviews the major developments. It starts with Cornfield's demonstration of odds ratio invariance under cohort versus case-control sampling, proceeds through the still-popular Mantel—Haenszel procedure and its extensions for dependent data, and highlights (conditional) likelihood methods for relative risk regression. Recent work on nested case-control, case-cohort, and two-stage case-control designs demonstrates the continuing impact of statistical thinking on epidemiology. The influence of R. A. Fisher's work on these developments is mentioned wherever possible. His objections to the drawing of causal conclusions from observational data on cigarette smoking and lung cancer are used to introduce the problems of measurement error and confounding bias. The resolution of such difficulties, whether by further development and implementation of randomized intervention trials or by causal analysis of observational data using graphical models containing latent variables, will challenge future generations of statisticians.

ISSN:0162-1459    

DOI:10.1080/01621459.1996.10476660

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

We determine factors, both hospital-specific and market area-specific, associated with hospitals' propensities for discharging Medicare stroke patients to skilled nursing facilities (SNF's) in California and Florida. Logistic regression is generalized to the case of a betabinomial, hierarchical model, in which covariate information is included in the hyperparameters of the second-stage beta distribution. It is found that the posterior mean of the proportion discharged to SNF is approximately a weighted average (i.e., shrinkage estimator) of the logistic regression estimator and the observed rate. We develop fully Bayesian inference that takes into account uncertainty about the hyperparameters, and we find that this also allows us to test for overdispersion in a natural way. The number of observed zeros (i.e., hospitals that sent no stroke patients to a SNF) is excessive compared to the number expected from a standard logistic regression model and is fit better by the hierarchical betabinomial model. The factors associated with discharge to SNF differ between California and Florida. In California the case-mix index and percent Medicaid admissions of the hospital, as well as the per capita income for the area and whether there is a rehabilitation facility in the area, are associated with discharge rates to SNF's. In Florida, whether there is a rehabilitation facility in the area is the only factor that exhibits association with discharge rates to SNF's.

ISSN:0162-1459    

DOI:10.1080/01621459.1996.10476661

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

A framework is proposed for the analysis of ordinal categorical data when ratings from several judges are available. I emphasize the tasks of estimating latent trait characteristics of individual items, regressing these latent traits on observed covariates, and comparing the performance of raters. The model is illustrated in the design and evaluation of an automated essay grader. This grader is based on a regression of variables, obtained from a grammar checker, on essay scores estimated from a panel of experts. The performance of the grader is evaluated relative to human graders, and implications on the reliability and repeatability of both automated and human raters is investigated.

ISSN:0162-1459    

DOI:10.1080/01621459.1996.10476662

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

This article develops methods for estimating treatment effects in mixed-effects models using outcome data gathered from serial dilution assays. Our application allows us to estimate the viral burden of HIV infection before and after antiviral treatment from cell dilution assays. This assay is designed to determine the infectious units per patient peripheral blood mononuclear cell (PBMC). The infectious unit is the amount of virus required to produce detectable HIV infection in PBMC's from healthy, uninfected donors. At each dilution level of the patient cells, one observes whether or not it was possible for the virus from these cells to infect donor cells. Thus the assay result for each subject consists of a series of repeated binary outcomes. We propose an analytic approach in which patient-specific titers (measures of viral burden) are modeled as random effects from an unknown distribution, and treatment effects are modeled as fixed. This approach makes use of all assay results, even if many assays fail to reach endpoint (i.e., they turn negative at the highest dilution level) and the assay design (dilution scheme) changes over time.

ISSN:0162-1459    

DOI:10.1080/01621459.1996.10476663

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

Pharmacokinetics is the study of the time course of a drug and its metabolites following its introduction into the body. Population pharmacokinetic studies are becoming increasingly important as an aid to drug development. The data from such studies typically consist of dose histories, drug concentrations with associated sampling times, and often covariate measurements such as the age and weight of each subject. These studies aim to provide an understanding of the pharmacokinetics of the drug in question and so lead to an informed choice of dosage regimen. Such an understanding includes determining those covariates that are important predictors of fundamental pharmacokinetic parameters, such as clearance, defined as the volume of plasma cleared of drug in a unit of time. Determining those subpopulations (e.g., the elderly) with altered kinetics has implications for the choice of an appropriate dosage regimens, because predictive concentration profiles arising from a particular regimen in different populations may be very different. In this article a general Bayesian hierarchical model is described. Pharmacokinetic models relating concentration to time are generally nonlinear, and the data are often sparse and/or noisy. The number of individuals on whom data have been collected is often large, and so the dimensionality of the parameter space is large. Consequently, estimation, from a Bayesian or a classical perspective, is not straightforward. In this article the Hastings—Metropolis algorithm is used for learning about the posterior distribution. An analysis of concentration data collected after the administration of the antiarrhythmic drug quinidine is presented. The data consist of 361 measurements on a total of 136 patients. Nine covariates are also available for each individual. These covariates are a mixture of discrete and continuous measurements. Some of the covariates are constant within an individual during the course of the study, whereas others change. A covariate model is constructed, and the sensitivity of the inferences to distributional assumptions is examined. The importance of assessing the appropriateness of modeling assumptions is emphasized and extensive model checking is carried out for the quinidine data using graphical diagnostics.

ISSN:0162-1459    

DOI:10.1080/01621459.1996.10476664

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

Given that all the cells of an individual have the same genetic information stored in their DNA, how can cells be as different as those of the retina and heart? Nature solves this problem through gene regulation, which often involves the binding of regulatory proteins to regulatory sites. These sites are short subsequences of 10 to 20 DNA base pairs whose pattern may be multinomially modeled. These sites usually occur “upstream” of the genes they regulate in a segment of a few hundred DNA base pairs called the promoter. But the positions of regulatory sites within promoters vary and are unobservable. This uncertainty in site position misaligns the data and renders the indices of the observations uncertain. Data with uncertain indices arise commonly in experimental biology whenever uncontrolled variability alters unobservable auxiliary identifying information. Current technology breaks the analysis of such data into two steps: alignment and analyses applied to the aligned data. This article proposes a methodology that combines these two steps and thus produces inferences that directly incorporate random alignment errors. The introduction of an index permutation indicator variable, which is treated as missing data, permits the formulation of these problems as novel finite mixtures. Using a missing information approach, we separate the likelihood into components representing variable uncertainty and index uncertainty. An EM algorithm to obtain the maximum likelihood estimates of the parameters for both of these components is also presented. Inferences specific to the index permutations stemming from index uncertainty are examined. An application to regulatory sites for a bacterial regulatory protein—cyclic adenosine monophosphate receptor protein (CRP)—is presented.

ISSN:0162-1459    

DOI:10.1080/01621459.1996.10476665

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

In the Lung Health Study (LHS), compliance with the use of inhaled medication was assessed at each follow-up visit both by self-report and by weighing the used medication canisters. One or both of these assessments were missing if the participant failed to attend the visit or to return all canisters. Approximately 30% of canister-weight data and 5% to 15% of self-report data were missing at different visits. We use Gibbs sampling with data augmentation and a multivariate Hastings update step to implement a Bayesian hierarchical model for LHS inhaler compliance. Incorporating individual-level random effects to account for correlations among repeated measures on the same participant, our model is a longitudinal extension of the Tobit models used in econometrics to deal with partially unobservable data. It enables (a) assessment of the relationships among visit attendance, canister return, self-reported compliance level, and canister weight compliance, and (b) determination of demographic, physiological, and behavioral predictors of compliance. In addition to addressing the estimation and prediction questions of substantive interest, we use sampling-based methods for covariate screening and model selection and investigate a range of informative priors on missing data.

ISSN:0162-1459    

DOI:10.1080/01621459.1996.10476666

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

In this article we describe a likelihood-based regression model appropriate for analyzing incomplete multivariate binary responses. We focus on “marginal models”; that is, models where the marginal mean or expectation of the binary response is related to a set of covariates. The association between the binary responses is modeled in terms ofconditionallog odds ratios. When the nonresponse mechanism isignorable, it is not necessary to specify a nonresponse model, and valid inferences can be obtained provided that the likelihood for the responses has been correctly specified. But when the nonresponse mechanism isnonignorable, valid inferences can only be obtained by incorporating a model for nonresponse. An unresolved issue with nonignorable models concerns the identifiability of the parameters. So far, no general and practically useful necessary and sufficient conditions for identifiability are available. Here we suggest some simple procedures for examining the identifiability status of nonignorable models when the response variable is discrete. Finally, we present results for an analysis of multiple informant data from the New Haven Child Survey and the Eastern Connecticut Child Survey.

ISSN:0162-1459    

DOI:10.1080/01621459.1996.10476667

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

In the Bayesian approach to model selection or hypothesis testing with models or hypotheses of differing dimensions, it is typically not possible to utilize standard noninformative (or default) prior distributions. This has led Bayesians to use conventional proper prior distributions or crude approximations to Bayes factors. In this article we introduce a new criterion called theintrinsic Bayes factor, which is fully automatic in the sense of requiring only standard noninformative priors for its computation and yet seems to correspond to very reasonable actual Bayes factors. The criterion can be used for nested or nonnested models and for multiple model comparison and prediction. From another perspective, the development suggests a general definition of a “reference prior” for model comparison.

ISSN:0162-1459    

DOI:10.1080/01621459.1996.10476668

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor