Kamen's quick clustering of participation data: A research note

Leyland Pitt ,Deon Nel

doi:10.4102/sajbm.v19i2.974

Abstract

This brief research note revisits a simple, but very useful clustering procedure developed by Kamen (1970), and illustrates its use in the clustering of attitudinal/perceptual data. For the purpose of illustrating the technique, perceptions of peer group participation in potentially corrupt situations in business were used as the data set. The mean responses, standard deviations and medians of 458 managers served as input for a correlation matrix from which the variables were clustered. The clusters formed by the analysis have been interpreted as 'The Insiders', 'Felons', 'Happy Holidays', 'The Fiddlers', 'A Bit on the Side', and 'The Innocents'. From the clusters identified it was evident that some situations were similar in nature. Quick clustering of the pilot study data is regarded as successful and could therefore lead to improved questionnaire design as well as the elimination of similar questions.

Highlights

In many application settings there is reason to believe that the set of objects under study can be clustered into subgroups that differ in meaningful ways
The objective in most cluster applications is to arrive at clusters of objects that display intra-cluster variation relative to the inter-cluster variation (Dillon & Goldstein, 1984: Ch 5)
Quick clustering of pilot study data should lead to improved questionnaire design, both in terms of question semantics and with regard to the length of questionnaires - a number of 'almost duplicated' scenarios, such as one of scenarios 7 or 14, or one of 2 or 9 could either be eliminated, or replaced with a scenario which addresses another dimension of the ethical spectrum, in the case of this study

Summary

Introduction

In many application settings there is reason to believe that the set of objects under study can be clustered into subgroups that differ in meaningful ways. The objective in most cluster applications is to arrive at clusters of objects that display intra-cluster variation relative to the inter-cluster variation (Dillon & Goldstein, 1984: Ch 5) It differs from other methods of classification, such as discriminant analysis, in that the number and characteristics of the groups are to be derived from the data and are not usually known prior to the analysis (Afifi & Clark, 1984: 379). The cluster analysis procedures available will entail too much additional time and effort, both from the point of view of data processing, and with regard to complexity This brief research note revisits a simple, but very useful clustering procedure developed by Kamen (1970), and illustrates its use in the clustering of attitudinal/perceptual data

The data set

SD Median

Scenario numbers

Conclusions