On the Problem of Two-Dimensional Error Scores: Measures and Analyses of Accuracy, Bias, and Consistency

Describing and analyzing error for one-dimensional performance tasks is fairly straightforward, but suggestions for describing and analyzing error for two-dimensional performance tasks (e.g., marksmanship) are quite problematic. Specifically, imposing an arbitrary axis onto the two-dimensional work space, along which traditional one-dimensional measures can be computed and analyzed, yields measures of accuracy, bias, and consistency that are entirely dependent upon the choice of axis. The present contribution offers new measures and methods for describing and analyzing data from two-dimensional performances. Unlike the resu1ts from previous suggestions, the approaches described herein yield results that are completely independent of the axes used to quantify the individual two-dimensional trials. These new approaches are strongly related to well-established methods for describing and analyzing error for one-dimensional tasks.