Drawing a double jeopardy line

Abstract

This technical note describes the process for drawing a double jeopardy line as it would be estimated by the Dirichlet model. The resultant graph is an x-y plot of penetration and purchase frequency for brands of different market share under a given set of category conditions. Previous analyses show double jeopardy to be an approximately linear effect but the extrapolation inherent in this approach demonstrates the nonlinearity of the effect as brands approach high levels of penetration.