STATISTICAL METHODS FOR EPIDEMIOLOGIC STUDY OF STROKE

Columbia University Health Sciences, New York NY

Investigators

Linked publications & trials

Abstract

The primary goal of our project is to develop and evaluate statistical methods for neurological studies. The secondary goal is to disseminate these methods by writing computer programs, specifically SAS Macros. The statistical problems to be investigated in this proposal arise from the NIH-funded projects in which the Principal Investigator has been participating. Although these problems have frequently arisen in other fields as well, satisfactory solutions have not, as of yet, been found. Specific Aim 1 is to develop methods for the Cox proportional hazards model when covariates are partially missing. Specifically, we propose to develop a likelihood-based imputation method with or without specifying functional form of the distribution for missing covariate under various missing mechanisms. Specific Aim 2 is to develop methods for handling missing data in repeated measures analysis under various missing mechanisms. Specific Aim 3 is to develop a joint estimating equation method for handling nonignorably missing outcomes in bivariate binary data arising from Northern Manhattan Stroke Study.