Data reduction of discrete responses: an application of cluster analysis

Abstract

One form of data reduction is cluster analysis which is used to separate detailed data into constituent groups. This paper illustrates a useful application of cluster analysis to the data reduction of detailed discrete responses of a type that are often found in large surveys. This method is illustrated using the ABS Training and Education Experience Survey 1993a (TEES). This survey provides a detailed range of information on individuals' computer skills. The information given in the survey is simplified by using a form of cluster analysis to combine categories of computer ability and skills. Wage equations are estimated using different assumptions on the degree of clustering.