[R] path analysis

[John Fox]

Dear Manuel,

Polychoric correlations imply only that the *latent* variables are
continuous -- the observed variables are ordered categories. Tetrachoric and
point-biserial correlations are special cases respectively of polychoric and
polyserial correlations. As long as you're willing to think of the
dichotomous variable as the dissection into two categories of a latent
continuous variable (and assuming multinormality of the latent variables),
you can use the approach that I suggested. This isn't logistic regression,
but it's similar to a probit model.

Regards,
 John

--------------------------------
John Fox
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 stat.math.ethz.ch 
> [mailto:r-help-bounces at stat.math.ethz.ch] On Behalf Of Manel Salamero
> Sent: Sunday, August 14, 2005 12:34 PM
> To: r-help at stat.math.ethz.ch
> Subject: Re: [R] path analysis
> 
> This solves part of my problem with the independent ordinal 
> variables, but my dependent variable is truly categorial 
> (illness/no illness). Polychoric correlation implies that 
> data are continuous, which in not the case. Is possible to 
> implement logistic regression in the path model?
> 
> Thanks,
> 
> Manel Salamero
> 
> ---------- Original Message ----------------------------------
> De: "John Fox" <jfox at mcmaster.ca>
> Data:  Sat, 13 Aug 2005 19:35:24 -0400
> 
> Dear Manel,
> 
> > -----Original Message-----
> > From: r-help-bounces at stat.math.ethz.ch 
> > [mailto:r-help-bounces at stat.math.ethz.ch] On Behalf Of 
> SALAMERO BARO, 
> > MANUEL
> > Sent: Saturday, August 13, 2005 2:02 PM
> > To: r-help at stat.math.ethz.ch
> > Subject: [R] path analysis
> > 
> > Someone knows if it is possible to perform a path analysis with sem 
> > package (or any other) to explain a dependent
> > *dichotomus* variable?
> > 
> 
> Yes -- you can use the hetcor() function in the polycor 
> package to generate
> a correlation matrix and boot.sem() in the sem package to get standard
> errors or confidence intervals. Make sure that the 
> dichotomous variables are
> represented as factors. See ?boot.sem for an example.
> 
> I hope this helps,
>  John
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! 
> http://www.R-project.org/posting-guide.html