A METHOD FOR MEASURING LATENT VARIABLES OF SOCIAL OBJECTS BY MEANS OF PATTERN RECOGNITION THEORY.

This paper deals with a task which is typical for sociological research. In the aspect of content it was stated in the following manner: supposing there is a set of occupations; sociologists can collect questionnaire data for analysis. The task is to scale these occupations according to the subjects' answers. The matter of scaling is in replacing the object designations by numbers in such a manner that the relationships on objects might be retained in numerical relations. The measuring procedure is in assigning numerical values to the objects, and it is executed by the third element of the scale, namely by the function. Depending on arbitrary definitions of the function, the following scale types are considered: absolute, interval, differential, ordinal and nominal. Thus, constructing a scale of designations would consist merely in replacing different names of occupations by different numbers. The representatives of the same occupation would not differ from each other. In constructing mathematical models such distinction is insufficient as the proportion of studies in which the results of nominal scaling are used is very small.