Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Re: Factor analysis using ICE


From   Maarten Buis <maartenlbuis@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: Re: Factor analysis using ICE
Date   Sun, 24 Jul 2011 17:11:00 +0200

On Sun, Jul 24, 2011 at 1:04 PM, Ari Dothan wrote privately:
> In an advice you gave in the
> Statalist, http://www.stata.com/statalist/archive/2007-12/msg00507.html, you
> suggest as follows: "Notice that it is not necesary to let -ice-
> make multiple imputed datasets if you are only interested in the point
> estimates for the factor scores. The multiple imputations are only used for
> adjusting the standard errors"
> I made 10 imputations using my dataset. They yield values which are a lot
> different from each other.
> Questions:
> 1. Don't I just select one random value of the imputed var. if I make just
> one imputation? Isn't it meaningless?
> 2. If the imputed value is negative in cases in which this is realistically
> impossible (such as negative assets) do I replace negative values with
> (missing)?

Please do not ask questions privately, but use the Statalist for that.
See: <http://www.stata.com/support/faqs/res/statalist.html#private>
for the reasons why.

In general you will get the right point estimates when using only one
imputation, but the argument is obviously asymptotic. So there can be
large variation form imputation to imputation. In that case you'll
probably be better of presenting the average over imputations.

However, the fact that you reported large differences between
imputations made me worried. There is some inherrit unidentifiebility
in factor analysis. If the identification depends on imputed values
(e.g. setting the variance of the latent factor at 1), than I doubt
whether you can meaningfully combine the estimates from different
imputations.

At this point I would advise against combining multiple imputation and
factor analysis, until you solved this problem of identification.

As to your second question, see
<http://www.stata.com/statalist/archive/2009-10/msg00861.html>

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany


http://www.maartenbuis.nl
--------------------------

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index