Acta Geodaetica, Geophysica et Montanistica Hungarica 24. (1989)
3-4. szám - Hajagos B.–Steiner F.: Measure of the linear dependence
LINEAR DEPENDENCE 419 applicability of Eq. (1) at the end of the present paper. 2. GENERATION OF A LINEAR DEPENDENCE In the following a simplified discussion is striven at according to Einstein's principle (which refers here both to the construction of the model and to the definition and the method of determination of the measure of dependence): "whatever you do, do it in the simplest way - but even simpler do not do it". 2.1 Family of distribution types f^Cx) Let us define the family of types (supermodel) f^(x) with the parameter oC (0 < oC < 2) as follows: » КЛИН-100 con. cotff.49(9 (conventional formula, Eq.1) согт. cotff. «Û0W (Eq.12) Xе • • Fig. 1. Random data pairs from a two-variable Gaussian distribution in case of independent x and y. One of the data pairs (0, 0) is burdened by a gross error. The traditional correlation coefficient is 0.989 Itl*/* Vx) = 7Г(2)