?
Корректировка погрешности вычисления коэффициентов корреляции при малых выборках
It is shown that the calculation of the expectations, standard deviations and coefficients of correlation gives significant errors when using small samples. Error calculating the correlation coefficient is significantly greater than the error calculation of the mathematical expectation and standard deviation. The cause of the error is quantized continuum source data through their representation of a small sample. We give the appearance probability distribution plots of quantization errors and errors arising in the calculation of the correlation coefficient in small samples. These values indicate the presence of significant methodological errors that arise when measuring the correlation coefficients. For small samples of systematic error is significant, but it decreases rapidly with an increase in volume of the test sample. It is proposed to correct systematic error in the additive and/or multiplicative form.
