Abstract:

The paper considers the current state of multivariate analysis when the data available only take the form of binary and classified variables. This restriction has meant that researchers have had to develop new techniques to deal with problems of defining and segmenting markets. An encouraging amount of work has been done in this field in the last twenty years and now there are statistical methods for handling the various types of dependence and interdependence analyses. The paper reviews some of the literature that shows how market analysts proceed with model building when the data are not normally distributed measured variables. Among the techniques cited are those under the general umbrella of Log Linear Model!ing, which enable regression and correlation type procedures to be employed and new ways of clustering respondents. The paper concludes that this growing area will continue to flourish especially as computer programs are available for many of the techniques.