Calculating correlated RV's with different stddev/var-coeeficient

[Effect of normalization on correlation matrix]

Hello dear list members,

Sorry for beeing confused about "subject" and "name" once more ;-(

For simulation purpose I need 2 RV's (X1, X2) with the same correlation
(r=0,3) to a target RV Y, same means (Y, X1, X2 = 1) but different stddev.
(X1 vs. X2 = 0,2 vs. 0,6)=> differing variation coefficients!!!

With "COMPUTE 0,3*Y+SQRT(1-(0,3 ** 2))*Y_1" i only get variables with r =
0,3 when the used Dummy-Var for calculating (Y_1) has the same
stddev/variation coefficient like my target-variable Y.

Here's my try. Does anyone know how to do that?

Greetings

Lenz


*Target-Variable.
*=============================================.
COMPUTE Y = RV.NORMAL(1,0.2) .

*Y_1,2,... = Dummy-RV for calculating RV's correlated with Y  (X1,
X2, ...).
*--------------------------------------------------------------------.
COMPUTE Y_1 = RV.NORMAL(1,0.2) .
COMPUTE Y_2 = RV.NORMAL(1,0.6) .

*calculating RV's correlated with Y  (X1, X2, ...), r = 0,3.
*-----------------------------------------------------------------------.
COMPUTE X1 = 0.3*Y+SQRT(1-(0.3 ** 2))*Y_1 .
COMPUTE X2 = 0.3*Y+SQRT(1-(0.3 ** 2))*Y_2 .
EXE.

[Marta García-Granero]

Hi Lenz

The process is a bit complex, and it's described in detail here:

http://support.spss.com/Tech/troubleshooting/ressearchdetail.asp?ID=19178

(use "Guest" both as user and password to login).

Since it uses MATRIX, if you run into troubles trying to adapt it to
your targets, tell me and I'll be glad to help (but the teacher in me
makes me hide the answer first in order to make you try to find it by
yourself, sorry!)

zbumd> For simulation purpose I need 2 RV's (X1, X2) with the same correlation
zbumd> (r=0,3) to a target RV Y, same means (Y, X1, X2 = 1) but different stddev.
zbumd> (X1 vs. X2 = 0,2 vs. 0,6)=> differing variation coefficients!!!

zbumd> With "COMPUTE 0,3*Y+SQRT(1-(0,3 ** 2))*Y_1" i only get variables with r =
zbumd> 0,3 when the used Dummy-Var for calculating (Y_1) has the same
zbumd> stddev/variation coefficient like my target-variable Y.

Regards,
Marta