Study of Political Preferences and Type 1 Errors in Traditional Correlation Approach

摘要

The article illustrates the problem of frequent absence of linear correlations in case of simultaneous simplest non-linear relations using the example of study of political preferences of young people that leads to system errors. The specific results are provided here to demonstrate type 1 errors when linear correlation shows the results close to zero. Such relations are simply not considered (even by supporters of "significant" correlations), but in fact there is often a real strong simple non-linear relation that completely changes the picture of a phenomenon or process under study. The results of data processing with the use of authorial method are presented for the case when the correlation analysis gives the result not exceeding the "significant" correlation, i.e. even the supporters of such errors leave these relations out of their view. All sufficiently strong dependences (the strength of relation coefficient 0.6 or more, a total of 26 dependences) are presented in the form of tables, and the quantile distributions for the dependences with the highest strength of relation coefficients (0.8 or more, a total of 13 dependences) are provided. Finally, we consider the sufficiently strong linear relation with a correlation coefficient of 0.6 or more.