An assessment of the minimum divergence criterion

The criterion of minimum divergence was first presented in this journal a few years ago as a tool from which an algorithm may be derived for the estimation of the state of a discrete linear dynamical system subject to noisy disturbances. Recently, the criterion was used as the basis of an algorithm for simultaneous state-parameter estimation. We show in this note that from both theoretical and practical viewpoints, the principle of minimum divergence is likely to lead to estimators with unnecessarily poor performance.