No, you can't add two R2 coefficients (let alone several), unless you are
absolutely certain that the predictor variables originating one coefficient
are totally independent from predictors in the other equation. One R2 tells
you what proportion of Y variance is explained by a linear combination of
Xs. Another R2 tells the same about another combination of Xs, which may or
may not contain some of the previous Xs, but is very likely to contain at
least some X that is correlated with some of the previous Xs. That is, some
of the variance explained by the second equation has been already explained
by the first, and therefore the two R2 partially overlap. The only surefire
way to know the overall proportion of variance explained by all the
predictors is to run an equation with all of them included (which may not be
always possible due to colinearity, excessive number of variables for your
sample size, or other reasons).
Hector
-----Mensaje original-----
De: SPSSX(r) Discussion [mailto:
[hidden email]] En nombre de
[hidden email]
Enviado el: Sunday, August 27, 2006 1:13 PM
Para:
[hidden email]
Asunto: variance explained
When obtaining correlation values and their respective resulting r2
(variance explained) can one or is it appropriate to ADD singular variance
explained
values to arrive at an additive or aggregate r2 even though a multiple
regression model containing the same predictors only results in an overall
adjusted
r2 that may or may not be greater or less than the aggregate singular r2s.
For example if I had the following coefficients, .17, .14, .12, .19 could I
add their r2 values to arrive at a total variance explained:
ca. 3% + ca. 2% + ca. 1% + ca. 4% = 10% of the variance explained.
I am aware of collinearity issues and understand that conceptual sense must
also be made of the results.
____________________________________________
Roland A. Carlstedt, Ph.D.
_www.americanboardofsportpsychology.org_
(
http://www.americanboardofsportpsychology.org)
[hidden email]
Locked
