ABSTRACT
Analysis of data which are qualitative in nature has become important since many of them originate from perceptions of people or from certain non-quantifiable attributes. This has wide applications now in social sciences, bio-sciences as well as behavioural sciences. For example, one may be interested to know the probability of survival of cancer patients depending on the stage of cancer, nature of the drug and age. Here the response variable is binary, i.e. whether the patient has survived or not. The explanatory variables are partly qualitative, like nature of the drug and stage of cancer. Sometimes in perception-based survey designs by researchers, the responses are ordered variables in terms of rank. The researcher is interested to know which type of socio-economic variable affects the specific ranking most. In these cases standard methods of quantitative analysis fails and one needs special techniques to estimate and analyse statistical relationships from the qualitative data. One such technique is logit (which uses a distribution known as logistic distribution for the variable to be explained) and another is called the probit (which uses the more common normal distribution). In both cases the variable to be explained is treated as discontinuous and binary having categories 1 and 0 only. The logit and probit use techniques to transform such discrete binary variables into continuous variables that are amenable to statistical estimations. This chapter discusses these techniques for analysing qualitative data.
