Dependencies in political studies (non-linearity and typical methodological errors)

Russian State Social University, psychological faculty, 129226 Wilhelm Pieck street, 4, build.1, Moscow, Russia

^* Corresponding author: basimov_@mail.ru

Abstract

The article raises the question of studying the statistical dependencies in a sociological research. Most sociologists, if they study cause and effect relations, offer interpretations of only linear relations and linear models based on their data. However, the problem arises not only with respect to the fact that sociologists interpreting any linear dependencies ignore a large number of simple non-linear relations (type 1 errors) often without understanding the essence of the issue. In the last 20-25 years, they often study processes not so simple to describe them by means of linear models. And sociologists go (consciously or not) along the way when weak linear relations (not strong ones are detected), referring to the hypothesis of zero correlation coefficient (saving stars SPSS), began to be presented as “significant” and tacitly understood as sufficiently strong correlations of a scientific interest for interpretation of cause and effect relations.But there is an even more significant error (type 2 errors), when they fail to notice not only the simplest non-linear dependencies, but strong simple non-linear dependencies between the parameters that provide their linear approximations with a weak correlation, and even a very weak correlation (0.11-0.3), which completely distorts the real picture of the phenomenon or process under study. It turns out to be scientific knowledge that does not correspond to reality, which contributes to the parallel development of philosophical (qualitative) analysis of social processes, based mainly on an intuitive understanding of social problems, the emergence of contradictions between the approaches. The article deals with some individual dependencies and their interpretation according to the results of the study of political preferences of young people, demonstrating the type 1 and 2 errors.