Relative error measures for evaluation of estimation algorithms

This paper is part of a series of publications that deal with evaluation of estimation algorithms. This series introduces and justifies a variety of metrics useful for evaluating various aspects of the performance of an estimation algorithm, among other things. This paper focuses on relative error measures, i.e., those with respect to some references, including the magnitude of the quantity to be estimated, its prior mean, and/or measurement error. It proposes several relative metrics that are particularly good for measuring different aspects of estimation performance. They often reveal the inherent error characteristics of an estimator better than widely used metrics of the absolute error. The metrics are illustrated via an example of target localization with radar measurements.