R help - May 2010 - SEM interaction

Hello all,

This is a general stats question--I realize it is an R help list, so 
tell me to go away if it is inappropriate.

I have a 2 X 2 design, and I have specified four identical path models 
(one for each level of each factor). I want to test for an interaction 
at each path--essentially (A1 - A2) - (B1 - B2) != 0. I was thinking of 
computing a contrast for each path of interest, such that I compute the 
difference of the difference in the (non-standardized) path weights (as 
above), and then divide this by the pooled standard errors of the 
estimates (average of the standard errors).

Does this seem statistically sound, or am I way off base? I typically 
compare path weights across two levels (e.g., age differences in a path 
weight for the same theoretical model) using the stacked model approach, 
but am not sure how to apply this to the interaction.

Anthony

-- 
Anthony Steven Dick, Ph.D.
Post-Doctoral Fellow
Human Neuroscience Laboratory
Department of Neurology
The University of Chicago
5841 S. Maryland Ave. MC-2030
Chicago, IL 60637
Phone: (773)-834-7770
Email: adick at uchicago.edu
Web: http://home.uchicago.edu/~adick/

Have you looked into multigroup analysis?  Is that what you're after?

If so, you can do multigroup analysis in lavaan fairly easily or, if you're
using the sem package, drop me a line.  I have some scripts that will do
multigroup analysis for an equal sample size.

See 
http://faculty.chass.ncsu.edu/garson/PA765/structur.htm#multiplegroupanalysis
and
http://www.structuralequations.com/3.html

for some examples.

On May 25, 2010, at 8:38 AM, Anthony Dick wrote:
> Hello all,
> 
> This is a general stats question--I realize it is an R help list, so tell
me to go away if it is inappropriate.
> 
> I have a 2 X 2 design, and I have specified four identical path models (one
for each level of each factor). I want to test for an interaction at each
path--essentially (A1 - A2) - (B1 - B2) != 0. I was thinking of computing a
contrast for each path of interest, such that I compute the difference of the
difference in the (non-standardized) path weights (as above), and then divide
this by the pooled standard errors of the estimates (average of the standard
errors).
> 
> Does this seem statistically sound, or am I way off base? I typically
compare path weights across two levels (e.g., age differences in a path weight
for the same theoretical model) using the stacked model approach, but am not
sure how to apply this to the interaction.
> 
> Anthony
> 
> -- 
> Anthony Steven Dick, Ph.D.
> Post-Doctoral Fellow
> Human Neuroscience Laboratory
> Department of Neurology
> The University of Chicago
> 5841 S. Maryland Ave. MC-2030
> Chicago, IL 60637
> Phone: (773)-834-7770
> Email: adick at uchicago.edu
> Web: http://home.uchicago.edu/~adick/
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.