Modern Interpretations of Correlations in Sociology As a PATH to Mock-Scientific Results

Tuesday, 17 July 2018
Distributed Paper
Mikhail BASIMOV, Russian State Social University, Russian Federation
In sociological science traces the exodus (intentional or subconscious) to low values of the correlation coefficient, when non-zero correlation (the hypothesis for correlation coefficient equality to zero) is sufficient to describe the statistical relations as strong.

This can be explained as a shortage of the really strong linear statistical relation with in magnitude higher 0.6 in the analysis, when studied sociological objects mostly non-linear in nature but used tools, still represented the linear models, and researchers do not want realize the synergy paradigm and non-linear models.

Let’s consider the model of exponential dependence within solving the problem of investigation of statistical relations using the author's method for 58 parameters: mathematical functions and regression lines for them, based on the correlation coefficient.

A dependency for a function with a single maximum (cut off on the right):

Y=-X^2 (left from the maximum);

Y=-0.7*X^2 (right from the maximum)

In this case, the correlation coefficient is equal to 0.25.

Dependence of the parameter Y from the parameter X as comparative weightiness of the parameter Y for quinters on a scale X:

X-1(Y=-12902); X-2(Y=-4658); X-3(Y=-742); X-4(Y=-3362); X-5(Y=-10978)

Dependence of the parameter Y’(regression line) from the parameter X:

X-1(Y’=-9616); X-2(Y’=-8328); X-3(Y’=-7196); X-4(Y’=-5730); X-5(Y’=-4866)

If we build graphics it will be apparent what a small part (21%) causation (and obviously one-sided) describes the regression line near the average value of the dependent parameter. While sociologists announce this correlation the «significant” and describe the dependence between parameters as linear.

A dependency for a function with a symmetric maximum:

Dependence of the parameter "Y=-X^2" from the parameter "X":

X-1(Y=-11944); X-2(Y=+6356); X-3(Y=+12310);

X-4(Y=+6356); X-5(Y=-11944)

Factor of the connection strength=1.65 (feedback is weak=0.00)

Coefficient of correlation=-0.00

The correlation is absent. In this case, is lost out of consideration not just the strong dependence but the dependence stronger than a linear function.