Estelle,
Just a question of clarification (my apologies if the answers seem obvious).
Did you use the regression weights from HALF1 to compute predicted Y1 in
HALF2? Did you use the regression weights from HALF2 to compute the
predicted Y2 in HALF1? You would have obtained these correlations if Y1 was
the predicted score based on a regression model that was built from the data
in HALF2, and Y2 was the predicted score based on a regression model that
was built from the data in HALF1. However, I do not want to assume that
this is how the analysis was carried out.
HTH,
Stephen Brand
For personalized and professional consultation in statistics and research
design, visit
www.statisticsdoc.com
-----Original Message-----
From: SPSSX(r) Discussion [mailto:
[hidden email]]On Behalf Of
Campenni, Estelle
Sent: Tuesday, February 06, 2007 12:09 PM
To:
[hidden email]
Subject: cross validation of regression output
In presenting double cross-validation to my graduate stat class, I found odd
coincidental results which I cannot explain.
Calculating regression on HALF1, R = .123
Using HALF2, correlating PREDICTED Y1 with ACTUAL Y, = .360
Calculating regression on HALF2, R = .360
Using HALF1, correlating PREDICTED Y2 with ACTUAL Y, = .123
I have done this numerous times but have not noticed this pattern before.
Any explanations?
Estelle Campenni
Marywood University
Locked
