How can i change the levels that categorical variabels are held constant in a Dummyregression

[jonassh]

Dear stata users

I am stuck with a problem that might be relatively simple for some of you, but is causing me some troubles

I am doing a regression with multiple independent dummy variables
The data is coming from 3 different districts, and I have 2 main dummy variables
making the matrix look like this

	district 1	district 2	district 3
dummy 1	0	1	1
dummy 2	0	0	1

When i'm doing my regression im interested in:.
- the effect on dummy1, when district changes from 1 to 2 and 3 combined.

i use a command like this:

reg yvar dummy1 dummy2

but my regression coefficient for dummy1 reflects the difference between district 1 and district 3. And what im interested in is seeing the difference between district 1 and district 2 & 3 combined.
though, dummy2 is not held constant?

So the question is how can i decide what level the others variables is held constant in, while interpretating my primary variable?

It is relevant for control-variables as well, that I want to hold constant at a certain level. is that possible?


hope i've made myself clear. otherwise I will gladly elaborate

[Brendan Halpin]

If you want to compare district 1 with the rest, then all you need to do is include dummy1 only. Effectively, you want to treat district as having only two categories, so a single dummy variable will suffice.

Note that your dummy variables as shown in the table are a little non-standard. Dummy variables take the value 1 when the original variable is in the relevant category, zero otherwise:

	district 1	district 2	district 3
dummy 1	1	0	0
dummy 2	0	1	0

In this case if you do
reg yvar dummy1 dummy2
the interpretation of the two dummies is the effect of being in category 1 (or category 2) relative to what's left out (category 3).

If you do
reg yvar dummy1
the interpretation of dummy1 is the effect of being in category 1 relative to what's left out (categories 2 and 3).

[jonassh]

ok, thanks

Another thing is, that it seems that at some points the other dummies are held constant on 0, while in oter occasions it is held constant at two
can i decide on which level it has to held constant?

[jonassh]

the thing about my two dummy variables is that they are independent of each other. its to unrelated variables.

but I think youre right about that I have to do a seperate model, only with the first dummy