Thanks for your further comment, which unfortunately makes
the matter less clear as far as I am concerned. I
thought you expected -factor- to yield stdp as a by-product;
my impression is that neither Stata 8 nor Stata 9
purports to do this. This could be simply that the programmers
stopped where they did -- as an occasional programmer myself
I have every sympathy -- or it could reflect a more
fundamental issue, namely that what you appeared to want is not
defined theoretically. Thus you didn't really answer my
question, so far as I can see.
But here the emphasis seems to be on producing factor
scores and then using them in a regression. That's a
different story altogether. You should have no Stata problem
in getting stdp results out of a regression model. I don't have
any creative thoughts in that territory, as that is something I
would never do.
Nick
[email protected]
Kenneth Greene
> Nick,
>
> Thanks for your reply. In this case, I'm using factor
> analysis to recover a
> latent ideology dimension and use it as a dependent variable
> in a regression
> model. Since the dependent variable is an estimate, I would
> further like to
> apply a WLS or FGLS model, for which I think I need the
> standard error of
> the predictions. My understanding is that score (in Stata 8)
> and predict
> (in Stata 9) generate factor scores through a regression
> technique -- so it
> seems plausible that I should be able to get post-estimation
> statistics like
> stdp. Any creative thoughts?
Nick Cox
> My reading of the help is that, syntax changes aside,
> Stata 9 won't help you get what you think you want
> any more than Stata 8. I have various difficulties
> understanding what people want from factor analysis
> and why they use it: in your case is there theory
> that produces standard error of prediction formulae
> for each set of factor scores, or are you just guessing
> by analogy?
>
> Kenneth Greene
>
> > Hi all: I would like to obtain the standard error of the
> > prediction (stdp)
> > for factor scores. My understanding is that factor scores
> > are generated
> > through a regression technique, so it seems as though I
> > should be able to
> > get stdp. However, when I use predict it tells me that the
> > last estimates
> > are not found. Am I doing anything wrong? Is there a
> > workaround I could
> > use? Also, I am using Stata 8 where factor scores are
> > obtained through the
> > score command. I understand that Stata 9 does this through
> > predict - does
> > that mean 9 will do what I want more easily? Thanks for your help.
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
