Hi all,
I am studying the relationship between two variables, y and x. I have
observations for 10 years, 51 states and 10 age groups. In my model I wanted
to introduce fixed effect by year, state and year-state interaction.
I run the following regression
xi:reg y x i.state i.year i.year*i.state
(then the identification comes from the variation across age groups)
When I run this regression I am getting results that I am not buying. In
order to check my results and given the fact that I am not interested in the
coefficients of the dummies, I tried to construct manually the fixed
effects.
I started with the model,
xi:reg y x i.state,
which it should be equivalent to the model,
reg dy_s dx_s, with
dy_s=y-my ; dx_s=x-mx; and mx=mean(x) by state
I run this two specifications and I am getting the same coefficients as I
expected.
I do the same for the model
xi:reg y x i.state i.year
which is equivalent to the model,
reg dy_sy dx_sy, with
dy_sy= dy_s - mdy_sy ; dx_sy= dx_s-mdx_sy; and mdx_sy=mean(dx_s) by year
and I getting the same results with both specifications. My problem comes
when I tried to replicate the model
xi:reg y x i.state i.year i.year*i.state
because by doing the triple d_mean, I get different results that are
different to the one that stata provides.
*
* 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/
