Core - Biostatistics

Zerbe, Gary

Abstract

The primary objective of the biostatistical core is to provide statistical support for all of the projects int he Center. Such support includes assistance in study design and appropriate sample size determination, data base management, summarizing, plotting, and analyzing data, and assistance in writing scientific papers and abstracts. Data analysis often involves estimating and comparing linear and nonlinear models fit to data collected repeatedly over time or drug dose on the same animals. This requires the use of advanced parametric and nonparametric analysis of variance techniques with which the statistical core is familiar, but which are not fully developed. The secondary objective of this core is to extent the statistical methodology available for the analysis of repeated measures and particularly longitudinal data from lesioned and grafted animals. The core's research effort will be focused on two areas. One is parametric; the other is nonparametric, requiring no distribution assumptions. More specifically, the aims for the research component of the core are first to combine univariate nonlinear random effects models with recently developed multivariate linear random effects models in order to derive a multivariate nonlinear random effects model capable of simultaneously estimating, comparing and correlating the growth rates and half lives of several nonlinear growth or decay curves, and second to develop a general linear model for longitudinal growth or time response curves based on randomization which encompasses a broad class of experimental designs and unifies inferences based on randomization as the standard general linear model does for inferences base on normal theory, and to apply this model to the problem of nonparametrically comparing growth curves after adjusting for covariates and possible informative censorship. These research topics are ripe for significant progress and of importance to many of the Center's project.