use lower-order residuals as predictors

[Bo]

Hi all,

I am working on using a higher-order model to predict outcomes. Instead of only examining the effect of the high-order factor, I am also interested in the effect of each lower-order factor factor controlling for the general factor. I was wondering whether it is possible to use both the general factor and all the lower-order factor residuals to predict the outcomes simultaneously. I read in some papers that we can do this in EQS (use residuals as predictors).

Conceptually, what I want is as follows. However, I was told that the model was not identified (actually as expected). I just do not know how EQS overcomes the identification problem.

Model.Hi<-'
X1=~x1+x2+x3+x4+x5+x6
X2=~x7+x8+x9+x10+x11+x12
X3=~x13+x14+x15+x16+x17+x18
X4=~x19+x20+x21+x22+x23+x24
G=~h1*X1+h2*X2+h3*X3+h4*X4
Y=~Y1+Y2+Y3+Y4
Y~G+c1*X1+c2*X2+c3*X3+c4*X4'

Best,

Bo

[Terrence Jorgensen]

I was wondering whether it is possible to use both the general factor and all the lower-order factor residuals to predict the outcomes simultaneously.

No, all paths would not be simultaneously identifiable.  If all Xs predict Y, there is no information left for the common-factor component of the Xs (G) to also predict Y, unless you have some reason to place equality constraints on the set of paths.

I read in some papers that we can do this in EQS (use residuals as predictors).... I just do not know how EQS overcomes the identification problem.

Nor do I.  Try asking on SEMNET

http://www2.gsu.edu/~mkteer/semnet.html

Terrence D. Jorgensen

Assistant Professor, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

http://www.uva.nl/profile/t.d.jorgensen