Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: RE: Negative Binomial Models


From   Jesper Sorensen <sorensen@mit.edu>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: RE: Negative Binomial Models
Date   Wed, 13 Oct 2004 20:07:20 -0400

I haven't read the Maddala, but my impression is that it is not a matter of should or should not, but rather that when the expected mean is sufficiently high, the poisson and neg binom start to look at lot like the normal distribution, and hence you are just as well off using OLS etc. Scott Long has a nice demonstration of this in his book on LDVs. Of course, having a couple of cases with high counts doesn't mean that the mean is high.

//Jesper




Should or should not?

I'd like to hear the grounds for that. I find it
difficult to believe that any such criterion tenet
aquam. You should use a negative binomial distribution
whenever fitted probabilities are close to observed
probabilities.

The underlying frequencies are not that germane.
But if the total number of individuals in the sample
is large, at least some cell frequencies are likely
to be. In fact, to get a good handle on probabilities
in the tail, a very large sample is a really good idea.

I'm interested in comments on this. If you type

. ssc type nbfit.hlp

there's a worked example in which the peak frequency
was 807.

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

-----Original Message-----
From: Nick Cox
Sent: 14 October 2004 00:38
To: 'statalist@hsphsun2.harvard.edu'
Subject: RE: Negative Binomial Models




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

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu]On Behalf Of MacDonald,
> John
> Sent: 14 October 2004 00:23
> To: statalist@hsphsun2.harvard.edu
> Subject: st: Negative Binomial Models
>
>
> Is there a maximum number of counts one should not exceed
> when using the negative binomial model? Even if there is a
> adjustment for exposure (e.g., rate of homicides in a city). > I have a data set with over 200 homicides in some cities and
> I saw a reference in an article of American Journal of
> Sociology to Maddala's 1977 econometrics textbook that said
> you should exceed 50 counts. Is this true?
>
>
>
> --------------------
>
> This email message is for the sole use of the intended
> recipient(s) and
> may contain privileged information. Any unauthorized review, use,
> disclosure or distribution is prohibited. If you are not
the intended
> recipient, please contact the sender by reply email and
> destroy all copies
> of the original message.
>
>
>
> *
> * 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