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

Re: st: System of Seemingly Unrelated Poisson Regressions (with Constraints)

From   Maarten buis <>
Subject   Re: st: System of Seemingly Unrelated Poisson Regressions (with Constraints)
Date   Sat, 7 Jun 2008 10:13:14 +0100 (BST)

--- Markus Lampe <> wrote:
> I am working with bilateral trade gravity models on a disaggregated
> level to find out if bilateral trade treaties which preferentially
> lowered tariffs on specific commodities enhanced respective trade
> flows. I used to do that with a quite simple system of log-linear
> sureg regressions like
> ln(bilateralimportcommodity1)=a1 + b1 ln(incomeimporter)+c1
> ln(distance)+d1 tariff-preference-dummy [and lots of control
> variables]
> ln(bilateralimportcommodity2)=a + b2 ln(incomeimporter) + c2
> ln(distance) + d2 tariff-preference-dummy [and lots of control
> variables]
> for several goods
> to estimate an "overall average effect" I restricted d1=d2=d3, etc.
> Now I got aware of a newer literature on gravity models that bothers
> about biases caused by "0" observations in the import matrix that are
> excluded by taking the logs. Santos Silva and Tenreyro, The Log of
> Gravity, Review of Economics and Statistics 88, 2006, 88(4), 641-658.
> I now that I might estimate poisson regressions in dependent variable
> levels and later use surest, but this does not allow me to impose
> constraints.
> Does anyone have any idea about the feasibility of my idea (a system
> of SU poisson regressions with constraints) or has ever tried
> something alike?

I am not sure whether you realy want to estimate poissons instead of
using a glm with the log link, because with poissons you make some
additional and heavy assumptions about the residual variance, while a
-glm- with log link can also take care of zeros in exactly the same way
without these additional assumptions. 

Anyhow the answer to your question is -suest-.

Hope this helps,

Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434

+31 20 5986715

Sent from Yahoo! Mail.
A Smarter Email
*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index