PIMS UNBC Distinguished Speaker Series: Linqun Wang

Many scientific research problems are concerned with relationships among a group of variables. In real data analysis often some of these variables cannot be directly or precisely measured, and instead indirect or imprecise measurements have to be used. Examples in health and environmental sciences include long-term systolic blood pressure, cholesterol level, drug concentration in patient's blood pressure, cholesterol level, drug concentration in patient's blood, exposure to air pollutants or radioactive substances. The measurement error problem arises also in econometrics, psychology and other social sciences. It is well-known that the naive statistical methods ignoring measurement error gives to biased estimates and therefore misleading conclusions. In this talk, I will first show some examples of measurement error and demonstrate its impact in simple linear regression models. I'll then give a brief survey of simple methods and techniques for correcting the measurement error effects. In particular, I will introduce the instrumental variable (IV) approach and apply this method to some real data examples.