--- Maarten buis <[email protected]> wrote:
> Another way of thinking about the difference between these two models
> is that you are modeling the geometric mean of wage, while I am
> modelling the arithmatic mean of wage.
Below you can see that the two models give largely similar results. So,
which one should you choose? There is a trade-off here:
On the one hand people in your discipline are apperently used to
-regress- in this context, but it is a pain to explain that the
coefficients you report are ratios of geometric means rather than
ratios of means. There is nothing wrong with geometric means and their
ratios other than that a lot of people aren't used to them, and start
getting worried about your analysis when there is no need to be
worried.
On the other hand, using -glm- will allow you to present your results
in terms of ratios of means, but apperently people in your discipline
aren't used to using -glm- in this contex. There is nothing wrong with
using -glm- in this context other than that a lot of people aren't used
to it, and start getting worried about your analysis when there is no
need to be worried.
Maybe the easiest way out is to add a paragraph to your methods section
where you discuss the difference between the two models. Say that you
choose one of these and say that you have also used the other model but
that the results where very similar.
*------------------------- begin example --------------------
sysuse nlsw88, clear
gen ttl_exp2 = ttl_exp^2
gen wage_log = ln(wage)
gen cons = 1
reg wage_log ttl_exp ttl_exp2 grade cons, eform("exp(b)") nocons
glm wage ttl_exp ttl_exp2 grade cons, eform link(log) nocons
*-------------------------- end example ----------------------
-- Maarten
-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands
visiting address:
Buitenveldertselaan 3 (Metropolitan), room N515
+31 20 5986715
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
*
* 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/
