The Analysis of Market Basket Models under Consideration of Marketing Mix Effects I

In marketing research, brand choice decisions on the disaggregated level are frequently modelled with a multinomial logit model. The analysis of brand choice decisions within single product categories is increasingly complemented by the simultaneous estimation of decisions in multiple categories. The results of the joint analysis of purchase decisions can be used for product range optimisation or goal-oriented direct marketing activities. Methodically, the explicit consideration of cross-category dependencies in the purchase decisions is focused on. Ignoring these category relations will lead to biased parameter estimates. Consideration of dependence structures requires the application of extended model approaches. Therefore, either techniques revealing correlations between categories, or methods considering explicit dependencies between categories in the model formulation are used. The explorative approaches based on data mining are adequate for data aggregation. The explanative approaches particularly concentrate on the identification and quantification of cross-category influences on the choice decision.
Several model approaches for estimation have been examined in this project previously. After a detailed examination of market basket model approaches, joint purchase analyses were conducted for modelling linked category purchase decisions based on a multivariate logit model and its extensions. Additionally, explorative and explanative model approaches were combined in a two-stage process. The nested logit approach has been chosen as a supplementary explanative model specification for modelling purchase and quantity decisions. Different specifications of the nested logit model have been carried out and their particularities in model estimation have been indicated.
In the further course of this project three objectives are in focus. First, the combination of explorative and explanative methods shall be further considered and their application in the field of market basket analysis is to be extended. In doing so, alternative price concepts in existing multivariate logit approaches are to be considered particularly.
Secondly, a combination of spatial statistics approaches with modelling joint purchases is to be implemented. The necessity for considering concepts from spatial statistics results from the dependence structures between the categories. It will be examined to what extent such a transfer is possible and how it can be realized.
Thirdly, the nested logit model shall be expanded by the influence of latent variables. This is necessary because purchase decisions are influenced by non-observable factors as well.

Principal Investigators
Boztug, Yasemin Dr.rer.pol. (Details) (Marketing)

Duration of Project
Start date: 08/2004
End date: 07/2006


Abe, M., Boztug, Y., Hildebrandt, L. (2004) Investigating the Competitive Assumption of Multinomial Logit Models of Brand Choice by Nonparametric Modeling, Computational Statistics 19(4), 635-657.

Bell, D., Boztug, Y. (2004) The Effect of Inventory on Purchase Incidence: Empirical Analysis of Opposing Forces of Storage and Consumption, in: C.A.S.E. Working Paper Series, No. 43, Humboldt-Universität zu Berlin.

Boztug, Y., Hildebrandt, L. (2005a) A Market Basket Analysis Based on the Multivariate MNL Model, in: 34th EMAC Proceedings, Rejuvenating marketing: contamination, innovation, integration.

Boztug, Y., Hildebrandt, L. (2005b) An Empirical Test of Theories of Price Valuation Using a Semiparametric Approach, Reference Prices, and Accounting for Heterogeneity, Discussion Paper 057, SFB 649, Humboldt-Universität zu Berlin.

Boztug, Y., Hildebrandt, L. (2005c) A Market Basket Analysis Conducted with a Multivariate Logit Model, Discussion Paper 028, SFB 649, Humboldt-Universität zu Berlin.

Boztug, Y., Hildebrandt, L. (2006a) Eine Simulation zum Test psychologischer Theorien der Preisbewertung, in: T. Posselt & C. Schade (Eds.), Quantitative Marketingforschung in Deutschland, Festschrift für Klaus Peter Kaas zum 65. Geburtstag, Berlin: Duncker & Humblot, 253-275.

Boztug, Y., Hildebrandt, L. (2006b) A market basket analysis conducted with a multivariate logit model, in W. Gaul & R. Kruse (Eds.), From data and information analysis to knowledge engineering, Heidelberg: Springer, in Druck.

Boztug, Y., Reutterer, T. (2006) A Combined Approach for Segment-Specific Analysis of Market Basket Data, Discussion Paper 006, SFB 649, Humboldt-Universität zu Berlin.

Boztug, Y., Silberhorn, N. (2006) Modellierungsansätze in der Warenkorbanalyse im Überblick, Working Paper, Humboldt-Universität zu Berlin.

Silberhorn, N., Boztug, Y. (2005) Modeling the impact of marketing activities on joint purchases with the nested logit approach, Working Paper, Humboldt-Universität zu Berlin.

Silberhorn, N., Boztug, Y., Hildebrandt, L. (2006) Estimation with the Nested Logit Model: Specifications and Software Particularities, Discussion Paper, SFB 649, Humboldt-Universität zu Berlin.

Silberhorn, N., Dannewald, T. (2006) Alternative Strukturen des Nested Logit Modells mit der PROC MDC, in: D. B. Pressmar & R. Bödeker (Eds.), Proceedings der 10. Konferenz der SAS-Anwender in Forschung und Entwicklung, Shaker, in Druck.

Last updated on 2020-10-03 at 16:42