Analysis of Differences in the Error Matrix
A major and required assumption of the error matrix is that the label from the reference data represents the “correct” label of the sample site, and that all differences between the map and the reference label are due to classification and/or delineation error. Registration differences between the reference data and the remotely sensed map classification are caused by delineation and/or digitizing errors. Data entry errors are common in any database project and can be controlled only through rigorous quality control. Analysis of the error matrix must include exploring how many of the matrix differences are the result of observers being unable to precisely distinguish between the map classes when the accuracy assessment site is on the margin between two or more map classes in the classification scheme. Analysis of the causes of differences in the error matrix can be one of the most important and interesting steps in the creation of a map from remotely sensed data.