Regression Calibration and Its Causal Variable Selection

Regression calibration is a statistical method used to correct for measurement error-caused bias when the analysis goal is to estimate the exposure-outcome association under a mismeasured continuous exposure. This talk will introduce two commonly used versions of this method first and will discuss, in a causal framework, how to include covariates in the calibration model and the outcome model for validity and for efficiency when using regression calibration to correct for measurement error-caused bias. The talk will also introduce R functions that implement the regression calibration method (no required reading).

Presenter: Molin Wang, PhD