MATCHED ANALYSIS IN CASE-CONTROL AND COHORT STUDIES

点击翻译按钮获取中文摘要

This study will investigate the validity and efficiency of neighborhood matching for case-control and cohort studies. The National Health and Examination Surveys I and II data were used in conjunction with neighborhood codes (i.e., specifying which individuals in the sample lived close together) to empirically determine the effect neighborhood matching would have upon validity and variance of estimates of risk of various conditions with respect to differing exposures. It was demonstrated that for some types of exposure-condition relationships, neighborhood matching was useful for controlling for confounding. However, there was a loss in efficiency due to a reduced number of matchable observations and a smaller number of degrees of freedom in the test statistics. These empirical examples can provide some guidance to researchers who contemplate neighborhood matching for an observational study. This project is one of the first known attempts of investigating the effect neighborhood matching has upon the analysis of observational data.