Re: st: RE: Factor analysis and dummy variables.

[adiallo5@worldbank.org]

Thanks to Nick and Mark for their inputs.

Mark, the polychoric and polychoricpca are running forever.
Is there a better way to do this?

Nick,
Unfortunatelly, he left for the weekend.

But basically, in his reasonning, the PCA is reserved for continuous
variables while the MCA is reserved for dummy variables. In the same
idea, categorical variables (more than two responses) are to be used
with PCA.

He is a statistician, I am not. I know very few of factor analysis. My little
understanding is that it creates a common factor from a dispersion in
varlist.

My problem is that I want to create an asset index.

factor radio tv frige bike moto car , pcf factor(5)
score wealthindex

xtile wealthgroup5 = wealthindex, nq(5)

The score I obtain after the factor is of poor quality since I do not have a
homogenous
distribution (i.e. approx 20%) of wealthgroup5 subgroups. I have the same
problem with
or without weights.

Any clue what is the best solution?

Many thanks.



                                                                                                                                           
                      "Nick Cox"                                                                                                           
                      <n.j.cox@durham.ac.uk>           To:       <statalist@hsphsun2.harvard.edu>                                          
                      Sent by:                         cc:                                                                                 
                      owner-statalist@hsphsun2.        Subject:  st: RE: Factor analysis and dummy variables.                              
                      harvard.edu                                                                                                          
                                                                                                                                           
                                                                                                                                           
                      12/10/2004 12:42 PM                                                                                                  
                      Please respond to                                                                                                    
                      statalist                                                                                                            
                                                                                                                                           
                                                                                                                                           
                                                                                                                                           
                                                                                                                                           




What did your colleague say was the
reason for it being "wrong"? What
are your criteria for recognising
a "best command"?

Nick
n.j.cox@durham.ac.uk

adiallo5@worldbank.org

> I know this question has been asked several times, but it's
> worth asking again.
>
> Is it right to perform PCA analysis with dummy (0/1) variables?
>
> I used to code :
>
> factor radio tv frige bike moto car , pcf factor(5)
>
> where radio tv ... are dummy (0/1),
>
> but a colleague just told me I am wrong.
>
> A MCA in his mind is better.
>
> What is the best command in that case? Just to make sure.
>

*
*   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/