Andrea -
Yours is indeed an interesting problem - I was hoping someone else on
the list might have a better idea, but as a starter, you could try
modeling n|N and N separately. For n given N, (only for those cases
where N > 0 )you could try GEE. In Stata this would be -xtgee- with a
binomial family and logit link to coount for the dependence within
firms. The command would look something like
xtgee n x1 x2 x3 if N>0,fam(bin N) link(logit) robust i(firm)
For N you could try a zero-inflated Poisson or negative binomial with
random effects (-xtpoisson- or -xtnbreg-)
Al Feiveson
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Andrea
Rispoli
Sent: Wednesday, December 31, 2008 8:44 AM
To: [email protected]
Subject: Re: st: regression with dependent variable ranging from 0 to 1
yes
On Tue, Dec 30, 2008 at 10:28 PM, Feiveson, Alan H. (JSC-SK311)
<[email protected]> wrote:
> Sorry, I missed the part about N being zero also . So the model has to
> take into account that N can be zero. As an example, n|N ~
> binomial(N,
> p) if N > 0, otherwise n=0; and N has a Poisson distribution. p and or
> the Poisson mean may vary randomly from firm to firm (random effects),
> but is fixed within a firm. May require a customized -ml- program, or
> maybe even a Bayesian approach. I am assuming that both N and n vary
> within the observations pertaining to the same firm. Is this correct?
>
> Al F.
>
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of Feiveson,
> Alan H. (JSC-SK311)
> Sent: Tuesday, December 30, 2008 3:19 PM
> To: [email protected]
> Subject: RE: st: regression with dependent variable ranging from 0 to
> 1
>
> Andrea -
>
> Are n and N counts? What range of values do they take on? Can n be
> thought of as a binomial sample from N? Then try -glm- as before but
> using n as the dependent variable, not G or H. Still, clustering on
> firms, should probably be taken into account. Perhaps this can be done
> using -xtlogit- on a 0-1 variable that sums to n.
>
> Al F.
>
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of Andrea
> Rispoli
> Sent: Tuesday, December 30, 2008 3:06 PM
> To: [email protected]
> Subject: Re: st: regression with dependent variable ranging from 0 to
> 1
>
> I typed:
> glm depvar indepvar1 indepvar2... indepvarn, link(logit) robust nolog
>
> when I tried
>
> glm depvar indepvar1 indepvar2... indepvarn, link(logit)
> family(binomial) robust nolog
> I got
> "note: depvar has noninteger values"
>
> Actually, the great majority of my data are zeros: 204581 over 213000
> observation. What do you mean exactly with "model separately"?
>
> Austin:
> Sorry, I gave a wrong information: this is actually 1-H, that is why I
> have several 0s Let me say it more in detail:
> Dep var = Log(1+G)
> where G=1-H and H=S(n/N)^2 (an Herfindahl index)
>
> I have 0 for all the cases in which H=1 and for all the cases in which
> n and N=0
>
>
> Al:
> no clustering variable, but indeed I have 10 observations for each
> firm (I use dummies for dealing with this) and indeed some indep
> variables have the same values for all the 10 obs and sometimes the
> depvar is the same for some of the 10 obs
>
> Thank you!
>
>
> On Tue, Dec 30, 2008 at 8:16 PM, Maarten buis
> <[email protected]>
> wrote:
>> --- Andrea Rispoli <[email protected]> wrote:
>>> It is an Herfindahl index of concentration, it ranges from 0 to 1
>>> (in
>>> principle) : in my specific case:
>>>
>>> Variable | Obs Mean Std. Dev. Min Max
>>>
>> -------------+-------------------------------------------------------
>> -------------+-
>>> H | 213620 .0190621 .0920916 0 .6477536
>>
>> How many zeros do you have? ( type in Stata: -count if H == float(0)-
>> ) Even though it is possible for a fractional logit to model a
>> dependent variable that includes zero (and one), if there are too
>> many
>
>> of these, then that might indicate that these zeros occur due to a
>> separate process and need to be modeled separately.
>>
>> -- Maarten
>>
>> -----------------------------------------
>> Maarten L. Buis
>> Department of Social Research Methodology Vrije Universiteit
>> Amsterdam
>
>> Boelelaan 1081
>> 1081 HV Amsterdam
>> The Netherlands
>>
>> visiting address:
>> Buitenveldertselaan 3 (Metropolitan), room N515
>>
>> +31 20 5986715
>>
>> http://home.fsw.vu.nl/m.buis/
>> -----------------------------------------
>>
>>
>>
>> *
>> * 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/
>
> *
> * 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/
>
*
* 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/
