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

[David Hoaglin <dchoaglin@gmail.com>]

Dear Suryadipta,

Thank you for Stata Tip #87.  As Maarten explains in the last
paragraph, the approach applies to all forms of multiplicative
effects.  For your continuous dependent variable, you could work out
interpretations along the lines of the odds ratios and odds in his
example.  They would not, however, involve odds ratios (or
incidence-rate ratios or hazard ratios).

I don't recall that your earlier messages mentioned the many zeros.
Your analysis should take that feature of the data into account.
Perhaps the model should apply only to pairs of countries that have
nonzero bilateral imports.

Regards,

David Hoaglin

On Mon, Jun 10, 2013 at 1:48 PM, Suryadipta Roy <sroy2138@gmail.com> wrote:
> Dear David,
> The dependent variable in my regression is indeed a continuous
> variable, highly skewed with a lot of zeros (bilateral imports between
> countries). I will refrain from using an odds ratio interpretation as
> you have suggested. Here is a link to Maarten's Stata tip # 87:
> http://www.maartenbuis.nl/publications/interactions.pdf
>
> 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/