Methodological problems and fundamental errors in the use of a correlation analysis on the example of political sociology

The article seeks to study the statistical dependencies in sociological research. The vast majority of sociologists studying the cause and effect relations, offer, on the basis of their data, an interpretation of only linear relations and linear models. But the problem arises not only when sociologists, interpreting the linear dependencies, ignore, often failing to understand the essence of the issue, a large number of non-linear relations having a simple form (errors of type 1). In the last 20 to 25 years the psychology has been mainly concerned with the study of not such simple processes that might fit into linear models. So, the sociologists take a path (knowingly or not so) in which weak linear relations (few strong ones are detected, if at all) with reference to the hypothesis that the correlation coefficient is equal to zero (the saving stars of SPSS) started being presented as “meaningful” and latently understood as fairly strong correlations that are scientifically interesting for interpreting the cause and effect relations. Actually, quite a few such correlations have been discovered, which made it possible to write long articles creating an illusion of prolific results. However, a still graver error lies here (errors of type 2) when the simplest non-linear dependencies are not just missed (dependencies with maximum and minimum, increasing and decreasing dependencies, considerably dissimilar from linear ones) while instead of strong and formally simple non-linear dependencies between the parameters what is supplied are their linear approximations with a weak and even very weak correlation (0.2–0.3), which completely distorts the real image of the phenomenon or process under study. Thus, scientific knowledge is produced that does not correspond to the reality, which fosters the parallel development of the philosophic (qualitative) analysis of social processes based mainly on the intuitive understanding of social issues. In order to overcome further distancing of qualitative and quantitative analysis in sociology, it is necessary to start studying, in the framework of most studies involving the research on the cause and effect relations, both linear and simplest non-linear dependencies.Работа выполнена при поддержке РФФИ, проект № 18-011-01071