Study of Political Preferences and Type 2 Errors in the Traditional Correlation Approach

Abstract

The complex nature of the subject of research was shown using the example of the study political preferences of young people, and the problem of “significant” correlations. An approach to errors that completely change the researcher’s worldview was considered. The specific results were shown demonstrating type 2 errors, when many researchers (psychologists and sociologists) present a weak but “significant” correlation as real linear model of a phenomenon or process under study, but in fact, a real strong simple non-linear relation is hidden behind it, which completely changes the picture of this phenomenon or process. Correlation analysis gives the result accepted in the research community as noteworthy when the correlation is “significant” (more than a critical value, “saving” asterisks SPSS). It is from this set of dependencies that we select only those for which the correlations are very weak (no more than 0.3), and the author’s relation coefficient shows that there is a strong, simplest non-linear relation. All dependencies are presented in the form of tables (the strength of relation coefficients of 0.6 and more, 26 dependencies) and the inverted distributions for dependencies with the greatest strength of relation coefficients (of 0.8 and more – 13 dependencies; of 0.7 to 0.8 – 5 dependencies; of 0.6 to 0.7 – 8 dependencies). Finally, we consider the sufficiently strong linear relation with a correlation coefficient of 0.6 or more (9 dependencies).