ABSTRACT
The purpose of this chapter is to examine statistical approaches that could be of interest for behavioural researchers. These techniques represent specialized multivariate approaches that have not yet been used extensively in behavioural research, namely (a) logit and probit models, (b) cluster analysis and (c) multidimensional scaling. For each technique, four basic questions are addressed: What is the basic purpose of the approach? What are the pros and cons? How does it work globally? How has this approach been used in past behavioural studies? Although these techniques have not been used extensively, it does not mean that they are not adequate or suitable for behavioural research. The numerous advantages of the techniques discussed throughout the chapter represent their potential to provide empirical evidence supporting various research objectives. However, their disadvantages constitute a source of concern that can be mitigated with a rigorous research design and an in-depth understanding of the approach. Logit and probit models are particularly useful when the researcher aims to predict an outcome or an event displaying a small number of values. Cluster analysis is appropriate when the researcher intends to identify groups of objects reflecting common alignment of elements. Multidimensional scaling is considered in order to transform individual judgments of similarity of preferences among accounting objects into distances represented in multidimensional space.
