PSYCHIATRIC EPIDEMIOLOGY: LONGITUDINAL STUDY METHODOLOGY

纵向研究通常被设计用来跟踪精神病学的过程

Longitudinal studies are often designed to follow the course of psychiatric patients over a number of years, so that long term outcome can be described and factors that may predict outcome can be identified. Longitudinal studies are also used in a variety of settings in the fields of psychiatric epidemiology and clinical psychiatric research. Indeed, the longitudinal study design is one of the most powerful research tools available to the modern psychiatric researcher. The goal of this research is to bring together a group of researchers with experience and expertise in longitudinal data methodology and in the analysis of psychiatric epidemiologic and clinical data. The overall objective is to study and extend currently available methods for the analysis of longitudinal data and to develop new parametric and nonparametric methods for signaling such longitudinal data. Longitudinal data which arise in the field of psychiatric epidemiology and clinical psychiatry are typically fraught with problems of incomplete or missing observations, weak measurement scales (ordinal or nominal scale variables), and an unequal number of observations in each comparison group. While there have been numerous advances in the biometric literature in recent years, much work remains to be done to strengthen the techniques available to the psychiatric epidemiologist and clinical researcher. To this end, we have identified four specific aims, each dealing with methods for the analysis of longitudinal data. These aims are: 1) to study and compare the properties of methods for analyzing longitudinal data when some of the data are incomplete (i.e., when there are missing observations); 2) to develop semiparametric methodology for analyzing repeated measures of an ordinal scale response variable; 3) to develop goodness-of-fit procedures for assessing the adequacy of parametric and semi-parametric regression models for longitudinal data; 4) to develop nonparametric methods for comparing groups of subjects when the response variable is non-normal. These aims will be pursued while simultaneously applying the results of these methods to existing psychiatric data sets. Particular emphasis will be given to the data sets which have arisen in the field of psychiatric epidemiology and in the clinical research studies being conducted as part of the University of Iowa's Mental Health Clinical Research Center.