|
Abstract:
|
Marketing scientists are regularly faced with scale measured items, from which segments have to be derived. It is commonly known that this type of data frequently exhibits scale usage patterns caused by individual or culture-specific response styles. Such data will produce differentiating but meaningless results, if these response biases are not taken into consideration. This paper presents an approach providing both pre-clustering diagnoses on the likely impact of scale usage effects on segmentation results and a way to overcome these effects. Other than conventional re-scaling or centering, the suggested approach addresses the scale usage issue analytically without modifying the original data. As an alternative to Bayesian model formulations (Rossi, Gilula & Allenby, 2000), the suggested approach is less computationally intense and easily applicable to mass data.
|
