Modelling purchase patterns on the basis of incomplete and biased consumer purchase diary and UPC panel data

Abstract

This study proposes a new approach for the analysis and prediction of product class and brand purchase patterns. The approach is designed to specifically consider data consisting of incomplete and biased purchase histories that may result from diary or UPC panel samples. Alternative approaches are developed that are illustrated on the basis of actual panel diary data and statistical estimation procedures to describe and predict actual product class and brand purchase patterns. A sensitivity analysis demonstrates the potentially substantial errors that may result from the analysis of panel data in practice. © 1984.