Re: st: Count data interaction terms

[Maarten Buis <maartenlbuis@gmail.com>]

On Tue, Aug 23, 2011 at 3:54 PM, DA Gibson <DAGibson1@sheffield.ac.uk> wrote:
> If i run a poisson analysis for number of doctor appointments with
> independent dummy variables (equalling 1) for being obese and smoking
> as well as an interaction term between them and get results showing
> IRR values of, for example, .3(baseline) 1.4(obese), 2.3(smoker) and
> 1.4(interaction).
>
> Then the interpretation is that an obese individual who doesnt smoke
> is likely to visit the doctor ((1.4-1)*100%) 40% more times than their
> non-obese counterpart. A non-obese individual who does smoke is likely
> to visit the doctor ((2.3-1)*100%) 130% more times than their
> non-smoker counterpart.

That is correct

> And the interaction term would suggest that an
> individual who both smokes and is obese would visit the doctor
> ((1.4-1)+(2.3-1)+(1.4-1)*100%) 210% more often than someone who isnt
> obese and doesnt smoke?

Not quite, obese and smoking respondents visits doctors 1.4*2.3*1.4 =
4.5 times (i.e. 350%) more often than their non-obese and non-smoking
counterparts.

In addition you can also look at the effect of being obese for smoking
individuals, which is 1.4*1.4= 1.96. So if you are a smoker than being
obese will lead to an 96% increase in the number of visits to the
doctor. Similarly the effect smoking when being obese is 2.3*1.4=
3.22, i.e. smoking leads to 222% increase in the number of visits to
doctors when a person is obese.

Finally it can be useful to interpret the interaction directly. In
this case being obese leads to a 40% increase in the effect of smoking
(and, as interaction effects are symmetrical, smoking leads to a 40%
increase in the effect of being obese)

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany


http://www.maartenbuis.nl
--------------------------
*
*   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/