Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


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

Re: st: Interpretation of interaction term in log linear (non linear) model


From   David Hoaglin <dchoaglin@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Interpretation of interaction term in log linear (non linear) model
Date   Thu, 13 Jun 2013 16:58:12 -0400

Dear Suryadipta,

Thank you for the additional details.

In many analyses of non-negative counts, where a Poisson distribution
is plausible, the presence of a substantial number of zero counts
leads to consideration of alternative models.  Sometimes a negative
binomial distribution is satisfactory.  If the number of zero counts
is large enough, further alternatives include zero-inflated models and
hurdle models.  Cameron and Trivedi (2013) have an extensive
discussion of various alternatives.  I have not seen your data, but
your mention of "many zeros" raises a warning flag.

"Following tradition" is sometimes a very weak justification for
choosing a particular approach to analysis.  It is essential to
understand whether the data are compatible with the "traditional
approach."  If not, the "traditional approach" is not valid, no matter
how popular it may be in the literature.

When I get a chance, I will try to look at the two papers that you mentioned.

Regards,

David Hoaglin

A. C. Cameron and P. K. Trivedi (2013).  Regression Analysis of Count
Data, second edition.  Cambridge University Press.

On Thu, Jun 13, 2013 at 3:23 PM, Suryadipta Roy <sroy2138@gmail.com> wrote:
> Dear David,
> Thank you very much for the wonderful suggestions! In the literature
> that I am working on, some of the important papers have already being
> applying Poisson maximum likelihood estimator to explain bilateral
> import (to deal with the problem of zero trade in the whole sample),
> and I am just following the tradition by applying it for my research
> question here (with some interesting differences in the results
> compared to the log-linear model). A couple of important works from
> the ones that I have come across are, e.g. "The log of gravity" by
> Santos Silva and Tenreryo (Review of Economics and Statistics, 2006),
> and "Estimating the gravity model with gravity using panel data" by
> Westerlund and Wilhelmsson (Applied Economics, 2011).
>
> Best regards,
> Suryadipta.
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index