Re: st: SEM & degrees of freedom

[Nick Cox <njcoxstata@gmail.com>]

Your previous posts got through, as a simple check of the archives would show

http://www.stata.com/statalist/archive/2013-12/msg00786.html

http://www.stata.com/statalist/archive/2013-12/msg00787.html

I can't comment on why no-one answered. See
http://www.stata.com/statalist/archive/2014-01/msg00055.html
for a reminder that Statalist is not a help line, so no-one is obliged
to reply. The same post underlines the spelling "Stata".

Nick
njcoxstata@gmail.com


On 7 January 2014 14:25, Daniel Coulombe <d_coulombe@sympatico.ca> wrote:
> Hello,
>
> This is the third time I am sending this question to the list. I have not
> yet received a single reply. I wonder if my message effectively get through
> the list. Can I get a confirmation that this is read?
>
> Can anybody tell me why the number of degrees of freedom is the same for the
> two following model specifications :
>
> 1: sem (Verbal@bv -> word similar) (NVerbal@bn -> matrices seq_quan)
> (Spatial@bs -> pattern r_design), covstruct(_lexogenous, diagonal) stand
> cov( NVerbal*Verbal Spatial*Verbal Spatial*NVerbal)
>
> 2: sem (Verbal -> word similar) (NVerbal -> matrices seq_quan) (Spatial ->
> pattern r_design), covstruct(_lexogenous, diagonal) stand cov(
> NVerbal*Verbal Spatial*Verbal Spatial*NVerbal)
>
> In model 1, I constrain to equality the coefficients for each factors, thus
> gaining 3 degrees of freedom over the unconstrained model 2. But when I run
> these, I get 6 degrees of freedom for BOTH models!
>
> LR test of model vs. saturated: chi2(6)   =     29.28, Prob > chi2 = 0.0001
> for model 1             DF should be equal to 9
> LR test of model vs. saturated: chi2(6)   =      9.76, Prob > chi2 = 0.1351
> for model 2             DF is correct (at 6)
>
> Do I oversee something, or there is a bug in STATA 13 SEM?
>
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/