ADAPTATION OF NEW STATISTICAL IDEAS FOR MEDICINE

点击翻译按钮获取中文摘要

DESCRIPTION: (Adapted from the applicant's abstract) The past thirty years have been a period of intense methodological development in statistics, with ideas such as proportional likelihood, robust estimation, jackknife/bootstrap methods, empirical Bayes techniques, longitudinal estimation, and EM algorithm becoming important. The long term purpose of this application is to shorten the transfer time between promising theoretical/methodological innovations and their biostatistical applications. This usually means linking the new methods to established statistical theory, and making clear their data- analytic advantages. the applicants' research is pursued from both the Stanford statistics department and medical school. Four areas are proposed here as focal points for forthcoming research: the use of complicance data in the analysis of clinical trials; bootstrap methods particularly as applied to bias-correction, metaanalysis, prediction, and the construction of confidence ellipsoids; Monte Carlo Markov chain estimators for the analysis of the type of contingency tables arising form genetic marker data; and practical formulas for simultaneous hypothesis testing, based on Hotelling's geometric arguments. Some of this work will be carried out in collaboration with biomedical researchers working on arthritis, AIDS, cardiovascular disease, genetics research, and other diseases in which compliance measurements are particularly useful.