Comparing main diagonal entries in normalized confusion matrices: a bootstrapping approach

When assessing map accuracy, it is typical to compare results from different classifiers that operated on the same image data set. While conditional kappa allows individual matrix categories to be analyzed with respect to either inclusion or exclusion error rates, conditional kappa is not used to compare individual matrix categories with respect to both rates concurrently. When this concurrent comparison is desired, the matrices are typically normalized and then examined on a cell-by-cell basis by inspection. While no parametric test of significance exists for such a cell-by-cell examination, sampling distributions for these main diagonal entries can be estimated by bootstrapping, allowing inferences to be made about the population. In this research, one procedure for assessing the statistical significance of normalized matrix cells on the main diagonal is described