Regressor Diagnostics for the Errors-in-Variables Model - An Application to the Health Effects of Pollution

Diagnostics designed to probe the biases in regressor coefficient estimates due to measurement error in the regressors are applied to two studies of the health effects of pollution exposure. The diagnostics indicate that inferences about the effects of pollution exposure are more sensitive to measurement error in the study with the more accurate measure of pollution exposure. It is shown how this finding illustrates an often overlooked theme about measurement error: if interest focuses on the sign of the coefficient of a specific regressor, it is measurement error in the other regressors that is of particular concern.