A Bayesian Model for Measurement Errors in Diagnosis of Rheumatoid Arthritis

We discuss effects of measurement error in the study of rheumatoid arthritis. The measurement error comes from two major sources: disease-specific Berkson error due to variations in the case definition, and individual specific classical error due to uncertainty in self-reported survey data. A mixture of classical and Berkson measurement error suggests an unobserved intermediate latent variable between true and observed diagnosis. We take a Bayesian approach using the Monte Carlo Markov Chain method to eliminate the missing latent variables. We then evaluate effect of measurement errors in modeling association between rheumatic arthritis and Carpal Tunnel Syndrome.