A Bayesian measurement model: reliable uncertainties and control over instrumental drift

We introduce a general Bayesian model for the analysis of measurements. The model uses an adaptive continuous autoregressive process to describe the instrumental error. Individual measurement uncertainties are also taken into account in addition to the observed sampling variance. Compared with conventional methods, better detection and control over instrumental error are achieved. Furthermore, the scatter in the inferred uncertainties is an order of magnitude smaller than that of the purely sampling-based uncertainties, which results in an improved reliability of the results obtained.