How can I include the coefficents for my baseline interaction?

[Soj Karimi]

Hello,

I am doing an linear probabilty regression and want to incude the coeffficents for the baseline (in this case educational level 3) in my outputtable. What I am basically testing is the effect of a treatment on health - but as I expect heterogeneous effects for different educational levels I need to check the interactions. When I do so I get the results for my treatment#educational levels 2 & 1 (as a comparison to educational level 3). For educational level 3 I get 0 & (.). I would like to also include the results for treatment#educational level 3 - is that possible? Basically the Idea is to say: the treatment with educational level 1 leads to x level of diseases. When educational level rises to educational level 2, the treatment leads to an additional x% increase in health ans so on.

The code:
regress health treat##ib3.education $controls, vce(cluster clusterlevel)
est sto est_health_01

Thanks in advance.

[Soj Karimi]

Not sure wheter I solved it correctley but I came up with the following solution:

regress health 1.treat##3.education 2.treat##3.education treat##ib3.education $controls, vce(cluster clusterlevel)
est sto est_health_01

--> Now the ommitted interactions between my treatment 1 and eductaional level 3 and my treatment 2 and educational level 3 are included.

[Carlo Lazzaro]

Soj:
to protect your resulst from the dummy trap (Dummy variable (statistics) - Wikipedia), Stata omits level(s) of categorical variable(s) (and unavoidably so).
The good news is that you can decide what to leave in via the -fvvarlist- options.

[Soj Karimi]

Thanks!

[Soj Karimi]

Soj:
to protect your resulst from the dummy trap (Dummy variable (statistics) - Wikipedia), Stata omits level(s) of categorical variable(s) (and unavoidably so).
The good news is that you can decide what to leave in via the -fvvarlist- options.

Dear Carlo,

thanks a lot for the response! Unfortunatley my solution frome above does not work. Due to colinearity many variables are ommitted. After discovering thatI tried the fvvarlist option: ibn. with the noconstant option to deal with colinearity. Still it doenst work for the interactions. Do you happen to know a soultion for that?

Right now i have this code:
regress health treat##ibn.education $controls, vce(cluster clusterlevel)
est sto est_health_01

--> Using that code all coefficents for education level 1,2, & 3 are shown. BUT the interactions for educational level 3 are still omitted.

meaning I want to heave:
treat1 x educ 1
treat1 x educ 2
treat1 x educ 3

treat2 x educ 1
treat2 x educ 2
treat2 x educ3

BUT I get
treat1 x educ 1
treat1 x educ 2
OMITTED

treat2 x educ 1
treat2 x educ 2
OMITTED

Can someone help?

Best,
Soj

[Soj Karimi]

Hello,

i think I figured it out. The problem was the double ##

Using soley one # solves the problem.

regress health treat#ibn.education $controls, vce(cluster clusterlevel)

Best,
Soj

[Carlo Lazzaro]

Soj:
yes, but it's hardly what you probably want.
Interaction without the main conditional terms (-treat- and -education-, in your case) are rarely used.
In your case, they do not fix your problem, but simply code something different.