CMStatistics 2020: Start Registration
View Submission - CMStatistics
Title: Thematic-accuracy quality control of slope and aspect classes based on an equivalence test Authors:  Virtudes Alba-Fernandez - University of Jaen (Spain) [presenting]
Francisco Javier Ariza-Lopez - University of Jaen (Spain)
Maria Dolores Jimenez-Gamero - Universidad de Sevilla (Spain)
Abstract: The thematic quality control of slope and aspect classes derived from a digital elevation model can be performed comparing the results on a product and a reference using a confusion matrix. Because of the nature of data (elevations and derived slopes and aspects), users can be more interested in checking proximity than equality toward the reference quality requirements. In the spirit of incorporating little or irrelevant deviations between the observed data and the target population in the definition of the hypothesis, a model equivalence test is proposed. Such deviations are assessed utilizing a $\phi$-divergence measure, which in turn can be consistently estimated. The asymptotic behavior of the resulting test statistic is studied, and a critical region based on the asymptotic null distribution is considered. The finite sample performance of the proposal has been studied through several simulation experiments. Finally, the new test is conducted over a real data set.