Colloquium- Erica E. M. Moodie - The SIMEX correction for measurement error: recent extensions for health research
TITLE: The SIMEX correction for measurement error: recent extensions for health research
SPEAKER: Erica. E. M. Moodie (McGill University)
ABSTRACT: The simulation-extrapolation (SIMEX) procedure for correcting measurement error was proposed more than 25 years ago by Cook & Stefanski. The approach needs no replicate data, but does require the analyst to know the distribution of the measure error. In this talk, I will explain how the SIMEX procedure works and why it was developed, and demonstrate considerations for its use in two-step procedures frequently encountered in causal inference where one first estimates a propensity score before using it to de-confound a treatment-outcome relationship. I will also discuss an extension to a setting where the measurement error is not mean zero, but rather the measured covariate systemically overestimates the true, underlying value.
Join Zoom Meeting
Meeting details to follow