Markov chain Monte Carlo estimation for the two-component model

The two-component model for measurement errors, which incorporates both additive and multiplicative disturbances into a regression model, has been found to accurately describe the observed heteroscedaseticity in analytical chemistry measurements. Estimation and inference using this model are difficult because of the presence of two unobserved errors in each observation. Markov chain Monte Carlo techniques enable exact inferences to be made for both the model parameters and unknown concentrations, but there are difficulties to overcome in achieving fast convergence to the target distribution. This article describes and illustrates an implementation of this method, and compares it with other estimation methods.