Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Re: Factor analysis after multiple imputation in STATA


From   tang5@purdue.edu
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Re: Factor analysis after multiple imputation in STATA
Date   Tue, 17 Jul 2007 10:32:33 -0400

Just a follow-up on this: The software Lisrel may have a way to solve the the 
OP's problem
 "Rodrigo A. Alfaro" <raalfaroa@gmail.com>:

> ///
> Woolton,
> 
> You could check Schafer, J.L. (1997), Analysis of Incomplete Multivariate 
> Data, New York: Chapman and Hall, as primary reference for Missing Data and 
> how to run EM. Joe has an executable for Windows that computes EM algorithm 
> (http://www.stat.psu.edu/~jls/norm203.exe), it is very fast and friendly. 
> You should take the variance-covariance matrix, and construct the 
> correlation matrix by hand (this is easy, it should be a shortcut... but I 
> am out of ideas), then use -factormat- to solve your problem.
> 
> I checked the web (long life to gooogle) and there are other sources for 
> your problem. Little and Rubin (2002) 
> http://www.amazon.com/Statistical-Analysis-Missing-Data-Second/dp/0471183865
> 
> and Jamshidian, M. (1997). An EM algorithm for ML factor analysis with 
> missing data. (Ed.). Berkane, M (Ed.). Latent Variable Modeling and 
> Applications to Causality (p. 247-258). New York: Springer Verlag. The 
> latter seems to what you want.
> 
> HTH, Rodrigo.
> 
> 
> 
> ----- Original Message ----- 
> From: "Woolton Lee" <finished07@gmail.com>
> To: <statalist@hsphsun2.harvard.edu>
> Sent: Thursday, July 12, 2007 8:58 PM
> Subject: Re: st: Re: Factor analysis after multiple imputation in STATA
> 
> 
> > Thanks for your reply.  So it sounds as though you are saying that I
> > can use the EM algorthm to run factor analysis in the presence of
> > missing data.  Is that correct?  How might I go about implementing
> > this approach?  By the way, do you have any references I might use to
> > get some background on this approach?
> >
> > Thank you for your help,
> >
> > Woolton
> >
> >
> > On 7/12/07, Rodrigo A. Alfaro <raalfaroa@gmail.com> wrote:
> >> ///
> >> It is not clear to me what you can get averaging the factor loading. For 
> >> a
> >> reasonable number of factors, you could print out your output for each
> >> imputed-dataset and try to find if there is any pattern. For example,
> >> variable x1 has some load between 0.7 and 0.74... if you get x2 has loads
> >> between -0.2 and 0.8, then I would think more carefully about the 
> >> imputation
> >> method.
> >>
> >> Alternative, you could try EM-algorithm to get the variance-covariance
> >> matrix under normality assumption. With that it would be clear how to 
> >> obtain
> >> the associated factors (there will be just one).
> >>
> >> Good luck, R
> >>
> >>
> >>
> >>
> >> ----- Original Message -----
> >> From: "Woolton Lee" <finished07@gmail.com>
> >> To: <statalist@hsphsun2.harvard.edu>
> >> Sent: Thursday, July 12, 2007 4:44 PM
> >> Subject: st: Factor analysis after multiple imputation in STATA
> >>
> >>
> >> >I am working with a dataset with many missing values across the
> >> > variables and have used multiple imputation via chained equations
> >> > (created by Patrick Royston) to generate 5 multiply imputed datasets
> >> > with the objective of running factor analysis to analyze the
> >> > relationships among the variables in the dataset.  However, it seems
> >> > that MICOMBINE is only tailored for regression type procedures and is
> >> > not appropriate for application when implementing factor analysis
> >> > after multiple imputation.  Is there a STATA command such as MICOMBINE
> >> > that can be used to obtain factor loadings from the multiply imputed
> >> > data or will I have to apply Rubin (1987) 's formula manually (via
> >> > MATA or programming) to obtain the factor loadings after running
> >> > factor analysis separately on each of the imputed datasets?
> >> >
> >> > As a side note, I think that Rubin (1987)'s formula applied to factor
> >> > analysis would simply be the mean of the of the factor loadings across
> >> > the imputed datasets (I have 5 imputed datasets) , but is this
> >> > correct, or should I be using a different formula for the factor
> >> > loadings across imputed datasets?
> >> >
> >> > I would greatly appreciate any assistance,
> >> >
> >> > Woolton
> >> > *
> >> > *   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/
> >>
> >> *
> >> *   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/
> >>
> > *
> > *   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/ 
> 
> *
> *   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/
> 


*
*   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/



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