R help - Apr 2008 - SEM with a categorical predictor variable

Hi,

we are trying to do structural equation modelling on R. However, one of our
predictor variables is categorical (smoker/nonsmoker). Now, if we want to
run the sem() command (from the sem library), we need to specify a
covariance matrix (cov). However, Pearson's correlation does not work on the
dichotomous variable, so instead we produced a covariance matrix using the
Spearman's (or Kendalls) correlation method, which works.

Running the sem() command on our model using that covariance matrix works
fine, but I am not sure if it was okay to make the covariance matrix using
Spearman or Kendall. Can we interpret the regression coefficients that we
find in summary(sem) just as if we had used Pearsons correlation in the
covariance matrix? Or is there any other way to define a SEM including
categorical variables without using a covariance matrix?

I appreciate every help. Thank you very much,
Vera


-- 
View this message in context:
http://www.nabble.com/SEM-with-a-categorical-predictor-variable-tp16425959p16425959.html
Sent from the R help mailing list archive at Nabble.com.

Dear Vera,

If the smoker/non-smoker variable is exogenous (as seems to be implied by
your calling it a "predictor") then you can simply create a 0/1 dummy
regressor for it and calculate covariances in the usual manner. The
coefficient for the variable would have the usual interpretation for a dummy
regressor. I wouldn't calculate correlations, as opposed to covariances,
since this would muddy the interpretation. A also wouldn't use Spearman or
Kendall correlations.

On the other hand, if smoker/non-smoker is endogenous, then you could use
biserial/tetrachoric correlations: see the hector() function in the polycor
package. There's an example (though with polytomous rather than dichotomous
variables) in ?boot.sem and in a paper on the sem package that appeared in
Structural Equation Modeling, which is available at
<http://socserv.mcmaster.ca/jfox/Misc/sem/SEM-paper.pdf>.

I hope this helps,
 John

--------------------------------
John Fox, Professor
Department of Sociology
McMaster University
Hamilton, Ontario, Canada L8S 4M4
905-525-9140x23604
http://socserv.mcmaster.ca/jfox

> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-
> project.org] On Behalf Of Lila86
> Sent: April-01-08 5:32 PM
> To: r-help at r-project.org
> Subject: [R] SEM with a categorical predictor variable
> 
> 
> Hi,
> 
> we are trying to do structural equation modelling on R. However, one of
> our
> predictor variables is categorical (smoker/nonsmoker). Now, if we want
> to
> run the sem() command (from the sem library), we need to specify a
> covariance matrix (cov). However, Pearson's correlation does not work
> on the
> dichotomous variable, so instead we produced a covariance matrix using
> the
> Spearman's (or Kendalls) correlation method, which works.
> 
> Running the sem() command on our model using that covariance matrix
> works
> fine, but I am not sure if it was okay to make the covariance matrix
> using
> Spearman or Kendall. Can we interpret the regression coefficients that
> we
> find in summary(sem) just as if we had used Pearsons correlation in the
> covariance matrix? Or is there any other way to define a SEM including
> categorical variables without using a covariance matrix?
> 
> I appreciate every help. Thank you very much,
> Vera
> 
> 
> --
> View this message in context: http://www.nabble.com/SEM-with-a-
> categorical-predictor-variable-tp16425959p16425959.html
> Sent from the R help mailing list archive at Nabble.com.
> 
> ______________________________________________
> 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.