Notice: On March 31, it was **announced** that Statalist is moving from an email list to a **forum**. The old list will shut down on April 23, and its replacement, **statalist.org** is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

From |
Joerg Luedicke <joerg.luedicke@gmail.com> |

To |
statalist@hsphsun2.harvard.edu |

Subject |
Re: st: negative binomial model in stata |

Date |
Wed, 2 Nov 2011 15:36:21 -0400 |

I am not sure if this can be answered easily without any context information. But generally and very broadly speaking, when using a varying intercept model (e.g., -xtnbreg- with the (default) -re- option) you allow the mean of your dependent variable to vary across the subjects that define your panels. This is not the case when you use a "pooled" model (e.g., -nbreg-). If it happens that the variance across panel units is exactly zero, the varying intercept model essentially reduces to the pooled model. However, if this variance is greater than zero, it usually makes sense to take that into account. Anyway, if this is news for you I would recommend reading some introductory material on panel data analysis and/or multilevel modeling before running models like these. A nice general take on multilevel modeling is: Gelman A & Hill J (2007): Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge University Press. More Stata specific is this one: Rabe-Hesketh S & Skrondal A (2008): Multilevel and Longitudinal Modeling Using Stata. Second Edition. Stata Press. And, more specific to count panel models, you could have a look at Chapter 14 of: Hilbe JM (2011): Negative Binomial Regression. Second Edition. Cambridge University Press. HTH, J. On Wed, Nov 2, 2011 at 2:23 PM, Jianhong Chen <jianhongchen1985@gmail.com> wrote: > Hi, Joerg > > Sorry for the citation. The reference is Allison, Paul D. and Richard > Waterman, 2002 “Fixed effects negative binomial regression models.” In > Ross M. Stolzenberg (ed.), Sociological Methodology 2002. Oxford: > Basil Blackwell. > > Thank you for your response. I will try to the compare the two > comands. I have another question. Do you think I can use this command: > nbreg DV IV controls year, cluster(Id) to deal with the panel data? > > Best, > > Jianhong > > > On Wed, Nov 2, 2011 at 2:06 PM, Joerg Luedicke <joerg.luedicke@gmail.com> wrote: >> >> If you cite something, you should provide the full reference. However, >> saying that using this model: >> >> xtgee DV IV controls year, family(nb) corr(sta4) robust >> >> instead of using this model: >> >> xtnbreg DV IV controls, re >> >> does not make a lot of sense as they are quite different and it >> depends on a lot of things which one to chose. The first one (the >> -xtgee-) will fit a population averaged model while the latter >> (-xtnbreg-) will fit a varying intercept model (aka random intercept >> or sometimes just called random effects model). You could obtain the >> same result that you get with running the -xtgee- model by using >> -xtnbreg-. Just try the following and compare the 2 outputs: >> >> xtnbreg DV IV controls year, pa corr(sta4) vce(robust) >> >> >> Joerg >> >> >> On Wed, Nov 2, 2011 at 1:04 PM, Jianhong Chen >> <jianhongchen1985@gmail.com> wrote: >> > Hi, everyone >> > >> > I am dealing with longtitudinal panel data using negative binomial >> > model. I think random effect would fit. Therefore, I use the following >> > command: >> > >> > xtnbreg DV IV controls, re. >> > >> > However, I read another paper which says that " xtnbreg is supposed to >> > estimate random and fixed effects negative binomial models, but I >> > don’t recommend either. Both of these models have a very peculiar >> > parameterization that does not do what it is supposed to do" and this >> > paper recommends use one of the following commands: >> > >> > nbreg DV IV controls year, cluster(Id) >> > xtgee DV IV controls year, family(nb) corr(sta4) robust >> > >> > What do you think is the best way to go? Thank you very much. >> > >> > Best, >> > >> > Jianhong >> > >> > * >> > * 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/

**References**:**st: negative binomial model in stata***From:*Jianhong Chen <jianhongchen1985@gmail.com>

**Re: st: negative binomial model in stata***From:*Joerg Luedicke <joerg.luedicke@gmail.com>

**Re: st: negative binomial model in stata***From:*Jianhong Chen <jianhongchen1985@gmail.com>

- Prev by Date:
**RE: st: New package -use10save9- available on SSC** - Next by Date:
**Re: st: RE: Dropping Duplicates that Aren't Exactly Duplicates** - Previous by thread:
**Re: st: negative binomial model in stata** - Next by thread:
**st: Robust Hausman test** - Index(es):