[R] Problem with package sem

John Fox jfox at mcmaster.ca
Mon Jul 10 22:44:40 CEST 2006

Dear Gang,

I'm not clear about the form of the model that you're trying to fit,
nor what the form of the likelihood would be for such as model, but the
short answer is that sem() requires that each covariance among observed
variables be for the same observations and hence for the same number
(N) of observations.

BTW, I didn't respond to your query to r-help because I was out of town
and unable to read email.


On Mon, 10 Jul 2006 11:36:06 -0400
 Gang Chen <gangchen at mail.nih.gov> wrote:
> Hi Dr. Fox,
> I posted the following message to the R e-mail list, but didn't get
>  any response. I was wondering whether you could shed some light on
> my  problem.
> Thank you very much.
> Gang
> ===
> From: 	  p_connolly at ihug.co.nz
> Subject: 	Re: [R] Package sem
> Date: 	July 9, 2006 5:49:41 PM EDT
> To: 	  gangchen at mail.nih.gov
> Hi,
> I am trying to run some path analysis with Dr. Fox's sem package. The
> number N in sem(ram, S, N) is supposed to be the total number of
> observations, right? However, in my situation the effective number of
> degrees of freedom for each observed variable is estimated by some
> auto-regression process, thus each variable bears a different DF. Is
> there a way I could run such a path analysis with a vector of DF's in
> sem?
> Thanks,
> Gang

John Fox
Department of Sociology
McMaster University
Hamilton, Ontario, Canada

More information about the R-help mailing list