A0832
Title: Matching quantiles estimation for discrete distribution
Authors: Hyungjun Lim - Korea University (Korea, South) [presenting]
Arlene Kyoung Hee Kim - Korea University (Korea, South)
Abstract: Analysis of independently collected data requires unpaired data analysis to account for the missing correspondence between the response variable and explanatory variables. Quantile matching estimation (QME) is one such method, which is built to find a linear combination of explanatory variables such that its distribution best matches the distribution of the response variable. Despite active research in the unpaired data analysis, no prior studies have been conducted for the case where the response variable follows a discrete distribution. We introduce a novel Poisson quantile matching estimation (PQME) as the first unpaired data analysis method designed for the discrete-count response variable. A simple yet effective algorithm of PQME is presented and its theoretical properties are proved. Simulation studies and real data applications are included to demonstrate both the practicality and the effectiveness of PQME compared to conventional methods such as GLM.
