Hi all,

My dependent variables are binary variables. I have two types of programs (Program A, Program B) and training.
I first created interactions between the type of program and the training

Code:
gen ProgAT= ProgA*training
gen ProgBT= ProgB*training
In my model, I include the interaction terms of the two programs with the training, the main effect for training is omitted since there should be no training when there are no programs available.

Code:
reghdfe f.women ProgA ProgB ProgAT ProgBT , absorb (id year) vce(robust)
Using -lincom-, I want to examine the effect of training with each program, on the share of women at managerial positions (the dependent variable).

My questions are -

1. Is this the right approach to examine the effect of training on the program?

Code:
lincom ProgA+ProgAT
lincom ProgB+ProgBT
2. I then want to examine the different effects between programs -

Code:
lincom ProgAT- ProgBT
3. When I ran the interactions the original variables interacted (i.e not the generated interaction variables), the results were different.
What could explain the difference?

Code:
reghdfe f.women 1.ProgA#i.training 1.ProgB#i.training, absorb (id year) vce(robust)

Many thank in advance!
G.

(Stata 13, Windows operating system)