Re: st: an estimation method question

[Xiang Ao <xao@hbs.edu>]

Thanks, Maarten, for the detailed description.  I probably did not 
describe our data very clearly.  There are different numbers of founders 
for each firm.  The smallest number is 2, it can be as many as 10 
co-founders.  As far as I understand, the number of categories need to 
be fixed in the methods you mentioned.  Is that right?
Thanks,

Xiang

Maarten buis wrote:
> --- On Tue, 16/3/10, Xiang Ao <xao@hbs.edu> wrote:
>   
> > We are trying to estimate an equation which has some
> > constraints by observation.  We are studying founders
> > of firms.  The dependent variable is the share of the
> > firm.  We are studying what factors influence the
> > shares.  We now have a problem:  for each firm,
> > the shares necessarily sum to one.  We were thinking of
> > dropping an observation per firm, but it turns out the
> > results are dependent on which observation to drop. 
> > Any suggestions on estimation methods would be helpful.
> >     
> With this type a data the key distinction is whether you
> have 2 or more than 2 categories whose proportions should
> add up to 1. In the former case you either use -betafit-
> or -glm varlist , link(logit) family(binomial) vce(robust)-.
> In the latter case you can use -dirifit- or -fmlogit-.
>
> you can install -betafit-, -dirifit-, and -fmlogit- by 
> typing in Stata:
> ssc install betafit
> ssc install dirifit
> ssc install fmlogit
>
> -betafit-, -glm-, and -dirifit- were discussed in this talk
> at the 2006 London Stata Users' Group meeting:
> http://ideas.repec.org/s/boc/usug06.html
>
> The -glm- trick is based on this paper:
> Papke, Leslie E. and Jeffrey M. Wooldridge. (1996) 
> "Econometric Methods for Fractional Response Variables with 
> an Application to 401(k) Plan Participation Rates". Journal 
> of Applied Econometrics, 11(6):619-632. 
>
> The -glm- trick is also discussed in the Stata tip:
> Christopher F. Baum (2008) "Stata tip 63: Modeling proportions"
> The Stata Jouranl, 8(2): 299--303.
> http://www.stata-journal.com/article.html?article=st0147
>
> The model implemented in -betafit- was discussed in a number 
> of papers:
> Ferrari, S.L.P. and Cribari-Neto, F. (2004). "Beta regression 
> for modelling rates and proportions". Journal of Applied 
> Statistics 31(7): 799-815.
> Paolino, P. (2001). "Maximum likelihood estimation of models 
> with beta-distributed dependent variables". Political Analysis, 
> 9(4): 325-346. 
>
> Smithson, M. and Verkuilen, J. (2006) "A better lemon squeezer? 
> Maximum likelihood regression with beta-distributed dependent 
> variables". Psychological Methods, 11(1): 54-71.
> -fmlogit- is basically a generalization of the -glm- trick to
> multiple categories.
>
> examples can be found here:
> http://www.maartenbuis.nl/software/betafit.html
> http://www.maartenbuis.nl/software/dirifit.html
> http://www.maartenbuis.nl/software/fmlogit.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/
>   
*
*   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/