A0893
Title: Density on density regression by spline mixtures
Authors: Yasumasa Matsuda - Tohoku University (Japan) [presenting]
Rei Iwafuchi - Tohoku University (Japan)
Abstract: The aim is to propose a regression analysis when both dependent and independent variables are density functions. Estimating a density by B-spline mixtures, a density is expressed as a vector whose components are given by probabilities of mixtures and introduces basic operations of summation, scaler multiple, and inner products to make a space of density functions a linear space. Theoretical justifications for spline mixtures are discussed. Collecting crime data in Tokyo together with human mobility by NTT DoCoMo, density on density regression is applied to them as empirical demonstrations.
