The Dirichlet Multinomial Distribution as a Magazine Exposure Model

Abstract
The authors examine methods by which the Dirichlet multinominal distribution (DMD) can be parameterized in the modeling of magazine media exposure distributions. The performance of alternative methods of parameter estimation of the DMD is compared with that of three other exposure distribution models. Performance is assessed in terms of accuracy of predicted versus observed distributions on 515 tabulated schedules derived from 1979 SMRB data. Results show the DMD is superior in performance to all other models tested, including the popular beta binomial distribution (BBD).

This publication has 12 references indexed in Scilit: