Hi Sam,
Setting a reference category is equivalent to omitting a category as an
independent variable. If you put in dem, rep, and ind id as independent
variable, then what's the substantive interpretation when you set all three
variables = 0? By omitting a variable (say ind), then you only include dem
and rep and so when you set them both equal zero, then the interpretation is
that the "person" is an ind.
So in running the regression, if you already have the dummies as variables,
just leave out which one(s) are the reference category from your regression.
If you have one variable for party id, which takes on values of dem, rep,
and ind, then you can do it by doing as.factor(), which splits the variable
into dummy variables and automatically leaves one out (unfortunately R
chooses which one to leave out, usually the lowest one if they are ordinally
ranked).
As for getting close to the estimates, you should get as close as you can
and try to figure out why you can't get exact estimates if possible.
2009/3/21 Sam Barrows <sbarrows at fas.harvard.edu>
Dear list....
We are seeking to replicate an ordered probit model. Amongst the
independent variables there are democratic id, republican id, independent
id, and also north central, south, west, northeast. All of these are dummy
variables. Below the table it states "The reference variable for party
identification is "Independent" and for region it is "North East".
In
running our regression we have not specified which variable is the reference
variable. Should we do so and, if so, do you know how?
Also, we can get close to the estimates, with the correct significance on
the different variables, but we can't get it quite there. Is it possible to
give some indication as to how close is close enough?
Thanks
Kyle & Sam
_______________________________________________
gov2001-l mailing list
gov2001-l at
lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
--
Patrick Lam
Department of Government and Institute for Quantitative Social Science,
Harvard University
http://www.people.fas.harvard.edu/~plam