Yes, there is a problem in suppressing the constant. Try running the
model on men and women separately--it may show you that the measure of
income you are using is a much better predictor for one than the
other. Note that you almost certainly have an endogeneity
issue--people who have more chronic diseases are more likely to have a
more sporadic work history, and lower income. The reverse direction
of causation is as least as plausible as the one you choose to model.
Can you use something more predetermined, such as educational
attainment at age 30, say, for older individuals, or parents' wealth,
etc.?
On 2/11/06, Rafael Terra de Menezes <[email protected]> wrote:
> I'm working with a count model (poisson) to show the relationship between
> health and income. My dependent variable is number of chronicle diseases of
> a person and the independent variables are genderand income, quite simple.
> I ran the regression and the LR test return awful results, and the
> coefficients were not significant. I tried to run the model without constant
> and the results were fine, including the poisgof test.
> Is there any problem in suppressing the constant?
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
