ttest over multiple subsamples

[Sa Fe]

Hi,
I have three variables, x, y, and z. x and y are categorical variables, x takes 0 and 1, but y takes 1, 2, and 3.

I want to do a t-test on z as follows:

(z if x==0 and y==1) == (z if x==0 and y==2)
(z if x==0 and y==3) == (z if x==0 and y==2)

(z if x==1 and y==1) == (z if x==1 and y==2)
(z if x==1 and y==3) == (z if x==1 and y==2)

How can I conduct such tests? I'd appreciate your help. Thanks,

[Leonardo Guizzetti]

One way is to model these in a single model.

regress z i.x i.y

[Sa Fe]

One way is to model these in a single model.

regress z i.x i.y

trying to figure out how to get if, for example, the following difference is significant?
(z if x==0 and y==1) - (z if x==0 and y==2)

[Leonardo Guizzetti]

trying to figure out how to get if, for example, the following difference is significant?
(z if x==0 and y==1) - (z if x==0 and y==2)

To keep it simple, you can run this as a t-test. Foor more formal testing, a combined model approach would be better but an intro background to regression would be useful.

ttest z if inlist(y, 1, 2) & x==0, by(y)