I am working with a count data model, where my dependent variable is
the number of children per family. The independent variables are
dummies (birth decades, occupational groups) and the data set is a
cross section.
Because there are uncertainties about the people who reported "zero"
children, I have to truncate the data at the zero values. As often
proposed, I used the Zero Truncated Poisson model for analyzing my
data, but the problem is that there is significant underdispersion
given.
Cameron and Trivedi (Microeconometrics Uisng STATA (2010), S. 575)
suggest for ordinary Poisson and significant Overdispersion, to use
the option "vce(robust" as to correct for this problem. Does this also
work for ztp and underdispersion?
Another possibility would be a Generalized Poisson Model, which
delivers smaller standard errors. But, if I compare the Akaike
Criterion for both if these models, the ztp gets a better value. Could
you give me a hint, which model I should use?
If I should use the Generalized Model, can you tell me, how I can get
the R squared?
Furthermore, I would like to calculate marginal effects, which I did
with "margins". Unfortunately margins is not supported by "outreg2".
Can I still use the mfx command in Stata 11 to calcualte the marginal
effect? (Because mfx is supported by outreg2).
