CMStatistics 2021: Start Registration
View Submission - CMStatistics
Title: Spatialised similarity analysis of two digital elevation models via a large number of two-sample multinomial problems Authors:  Francisco Javier Ariza-Lopez - University of Jaen (Spain)
Maria Dolores Jimenez-Gamero - Universidad de Sevilla (Spain)
Virtudes Alba-Fernandez - University of Jaen (Spain) [presenting]
Abstract: A Digital Elevation Model (DEM) is a bare earth elevation model representing the surface of the Earth. From the elevation model, other models can be derived, such as slopes, orientations, insolation, drainage networks or visual and watershed analysis. Usually, the comparison of two DEMs covering the same area is made by analysing the discrepancy between the two elevation models; however, we consider that a DEM is more than just elevations and that it also has spatialisation. Therefore, to make the comparison between two DEMs, a multivariate and spatialized perspective is proposed. Both the considered variables (categorized) and the way of spatialisation (natural or artificial) can be decided by the user. Motivated by this open question, the problem of testing a large number, say $k$, of multinomial two-sample problems is tackled. It will be assumed that the available data consist of $k$ pairs of independent samples, which may be bounded or increase with $k$, but at a lower rate than $k$. A test statistic based on the Euclidean distance between the estimated proportions in each pair of populations is designed. The null distribution of the test statistic is unknown, and its asymptotic null distribution is derived. The asymptotic power is also studied. A simulation study is carried out to evaluate the behaviour of the proposal in a wide range of scenarios. Finally, a practical application is introduced using two DEM datasets from the same zone (Allo, Navarra, Spain).