October 29, 2010
Gillian Heller
Department of Statistics
Macquarie University
Title:  Random sums regression - which level do we analyse?

Asbtract:  We consider regression models in which the response is the sum of a random number of terms. Typically data are observed at varying levels of detail, and this determines the method of analysis. A framework of models is developed, which takes into account the observation level and distributional assumptions on the components of the random sum. However the question arises as to whether estimation should be carried out on the sum, or on its components, in the situation where the components are observed but interest is focussed on the sum. Examples of application in hydrology and insurance are discussed.