very baby question

[Mahnoor Sohail]

sorry im panicking actually

so my results are :

from first stage got predicted height for children

second stage (probit model)
schoolenroll(0/1)= -0.60*male+0.23*heightforage -0.24(male*heightforageinteraction) (all are strongly significant)

1...now my supervisor asked me to do this separately for boys and girls, to get rid on the interaction term(to make interpretation easy)
so i regressed separately
now
for females the height for age stays positive but becomes insignificant
for males stays positive and stays significant

2....BUT when i do margins,dydx(heightforage) at (male=(1 0)) of my main interaction model
it reverses
for male it becomes insignificant
and female stays significant

can someone pls tell why are they opposing

[FernandoRios]

Hi Mahnoor,
if you provide the exact line of code and output that you get, it will be easier for anyone to give you advice regarding to what may be happening.
Best