Statalist


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

Re: st: factor analysis - interpreting the results, inclusion of dp variable


From   "georg wernicke" <georg.wernicke@googlemail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: factor analysis - interpreting the results, inclusion of dp variable
Date   Wed, 15 Aug 2007 15:12:26 +0100

hey nicola,

i guess it does not make much sence to use factoranalysis here. one of
my supervisors was pointing me on that and i did not spend enough time
on reviewing when it is reasonable to use since my thesis is due soon
(i guess blindly trusting somebody is not the best idea either...).
yes, i meant using predict by saying i stored them.

i was looking at the correlation matrix before, nothing obvious or
seriously correlated has shown up. when using coldiag2, two variables
could be identified as correlated. i was looking for ways to get my
model smaller, e.g. not to use 15 variables which were suggested as
pivotal in the literature. i am probably going to report a number of
tables where i include different variables.

thanks again for your help

georg

On 8/14/07, nicola.baldini2@unibo.it <nicola.baldini2@unibo.it> wrote:
> Sorry for jumping in the discussion, but did you have a look at the correlation matrix before spending times on complicated tests, like those available through -coldiag2- and -collin-???
> I obviously agree with Scott: don't include the dependent variable in the factor analysis. Also: which kind of variables do you have? While I used a factor analysis with constructs (e.g. perceptions on Likert scales), I would never use it with """real""" things (e.g. profits, patents, age, etc...), it is often too complicated to interpret the findings.
> P.S. Just to be sure: when you said "i stored the 4 factors as new variables", did you mean that you saved the factor scores with post-estimation command -predict-, right?
> Nicola
>
> At 02.33 14/08/2007 -0400, "georg wernicke" wrote:
> >hello stata list,
> >
> >i have a probably rather simple questions to which i cannot find good
> >links in the net.
> >
> >i have quite a number of independent variables which have been
> >identified as significant in the literature before (around 16). when i
> >add all them in a regression, not one single variable is significant.
> >i want to employ factor analysis to get a persimonious model.
> >
> >here are my questions:
> > 1. do i include the dependent  variable in the equation? (fac
> >dependent  independent1..... etc.)
> >2. the analysis is retains 4 factors and gives me a factor loading
> >table which i rotated. now what do i do with the rotated table? i
> >stored the 4 factors as new variables..do i simply ass them in the
> >regression equation or do i only run a regression with these factors
> >now?
> >3. or does the result of 4 retained factors tell me that my model
> >should consist of 4 independent variables only?
> >
> >sorry, i know some of the questions are probably quite blunt...but
> >still, i am lost on what to do with the results..
> >
> >thanks
> *
> *   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