A distance correlation index of spatial dependence for compositional data

Geographical data in economic, social or environmental sciences are usually recorded as compositions, i.e. relative frequencies, and a common inquiring problem concerns the analysis of these data over different geographical regions. In the present paper we define a new statistical test to verify spatial dependence of such geographical distributions based on distance correlation, a recently introduced measure of dependence between random vectors. The proposed index computes the non-linear spatial distance between distributions and can be applied on compositional frequency distributions. Simulations and an application on Italian electoral data are presented, to illustrate the capabilities of the proposed test to detect spatial dependence. © 2019 The Author(s). Papers in Regional Science © 2019 RSAI