need some help with interaction of two variables

#1
Hi everyone!
I will try to explain what my question is so that you can hopefully help me to find an answer.
Ok, I have, among much other information, data on mother education and mother employment status (once discrete and the other categorical)
Lets say, I want to predict children achievement (continuous variable) on the interaction of mother education and mother job. That is, even if the mother is more educated, it is well possible that she spends her time out of the home because she has a job and, as a consequence, the child may not receive the support needed.
How would you go about it?


Would you use a simple if condition, that is, if mother is educated BUT she works full time or would you build a dummy for that or would you build an interaction variable?
In this last case, how would you go about it? I mean would if I let mother education interact with her job, how would I then explain the results? As there are plenty of values for both mother education and her job..


any help would be really much appreciated!

thanks!
 

bukharin

RoboStataRaptor
#2
I would use an interaction term, then use -margins- to examine child achievement for different levels of maternal job + education. It depends on how many jobs you have though - if there are (eg) hundreds then this won't be easy to interpret, and you should consider combining the job categories.
 
#3
ok. What if i regress the child achievement on mother level of education if occupation status==1, 2 and so on so forth?
because yes, i have 7 codes for education and 7 for occupations!
 

bukharin

RoboStataRaptor
#4
Oh I see. Well unless you have a pretty huge dataset that's going to be quite overfit. You may want to collapse categories, eg maternal employment could become none, part-time or full-time.