Change reference category of dependent variable

[Don Vicendese]

Hi,

I am running the following code

mlogit cq25 i.group [pweight = cq25pp], rrr

The dependent variable has three levels. It all works fine except Stata is choosing level 2 as the reference level for the dependent variable when I want level 1 to be the reference level.
I think Stata is doing this as level 1 has n = 80 and level 2 has n = 201 - a default for the larger n I guess.

I know independent variable reference levels can be changed with ib# but how can the dependent variable's reference level be chosen?

Don

[Clyde Schechter]

mlogit cq25 i.group [pweight = cq25pp], rrr baseoutcome(1)

[Richard Williams]

If I ruled the world, the baseoutcome would default to 1, and picking the category with the highest frequency instead would be an option. I don't like using highest frequency as the default because different samples or subsamples could have different base categories, making it harder to compare results. Also, even if the categories aren't ordered, there is often a reason categories are numbered the way they are. For example, you will sometimes use mlogit with an ordinal dependent variable because the assumptions of the ordered logit model are violated.

[Don Vicendese]

Thanks very much Clyde and Richard for taking the time to respond. That code worked fine. In retrospect, it is pretty basic but I had made a search but couldn't find anything. Unhelpful search terms I guess. I kept getting posts about changing the reference level for a covariate instead of the outcome. Many thanks again,

Don

[Clyde Schechter]

Well, before you go on internet searches, or even searches of this forum, it always pays to check the help files and PDF documentation that is installed with your Stata. The command that was not working as you wished is -mlogit-. Looking at -help mlogit- you will find -baselevels- and -nobaselevels- among the options explained there.

[Don Vicendese]

OK- thanks again Clyde.