Outliers in Evidential C-Means: An Empirical Exploration Using Survey Data on Organizational Social Capital

Evidential C-Means (ECM) is a technique for cluster analysis, which has a methodology based on the Dempster-Shafer theory of evidence (DST). To date this technique has been theoretically discussed but has had limited application. Based on DST, ECM facilitates the association of objects to sets of clusters, rather than simply a single cluster. One feature of ECM is the facility for classifying cases to no cluster, the level of which is effected by the parameters in ECM (in particular , which controls for the datapoints considered outliers). In this study, the substantive effects of varying  are explored by investigating the relationship between organziational social capital and employee engagement. Drawing on a large-N survey of senior public sector executives, the clustering of different dimensions of organizational social capital is undertaken, and the relationship between those clusters and employee engagement analysed at varying levels of . The implications of the findings are discussed.