Estimating the Parameters of a Convolution

Abstract

Parametric estimation problems are considered when one observes a random variable which is the sum of two independent non-identically distributed random variables. Some basic identifiability theorems are given which suggest the method of moments as a general method for attacking such problems, particularly when one of the summand variables is discrete. Moment estimators and their asymptotic distributions are examined for various specific examples. Some applications are suggested.