Beruflich Dokumente
Kultur Dokumente
A description of the correlation coefficient r in the parts and then fully applies to
the relationship between X and Y and does not depend on the assumptions that apply to
the variables X and Y. Where now for the X and Y there are certain patterns, such as
how X and Y are distributed, then the correlation analysis could go further, among others,
determine the interval estimation and hypothesis testing. For this, assumed that X and Y
are normally combined distributed with two variables in which among others contains the
parameter ρ (read: rho) as a correlation coefficient.
From the normal population this that has two variables take all random samples of
size n and then calculate the correlation coefficients of r with the formula that has been
described above. Then obtained a set of correlation coefficient, r, r 2, r 3,. . . . . . . . . From
this collection we can form the sampling distribution of correlation coefficients and
subsequently the average and standard deviation can be calculated. The Average and
standard deviation for the sampling distribution of correlation coefficients will be given
the symbol µ r and σ r.
If all that random samples from normal populations with variable two ρ = 0, then
the sampling distribution of correlation coefficients be symmetric with µ r = 0. if it is
formed statistics:
r n-2
t =
1- r2
If the population from which random samples were taken have ρ = ρ 0 ≠ 0, then the
sampling distribution of correlation coefficients are not symmetric. In this case, using a
transformation will not cause a symmetric distribution that approaches a normal
distribution. Transformation that is used is Fisher's transformation:
1 1 + r
Z = ln
2 1- r
With this transformation, the normal distribution that occurred (a form of approach) has
an average and standard deviation:
1 + ρ0
µz = 1 2 ln
1 - ρ0
1
σ =
n -3
1 + ρ0
µ z = (1,1513 ) log
1 − ρ0