Dear Statalist user and experts,
I used a probit command to run just simple probit regression. And after fitting the model I wanted to obtain the average probability by gender.
Let’s assume the following model fitted using probit technique:
probit Unemp Female G2534 G3544 G4554 G5564 Mstatus Training Secondary University MAPhD
where Unemp is a binary take 1 if the individual unemployed and 0 otherwise. Female is a binary take 1 if female and 0 otherwise. G2534 is a binary take the value 1 if the individual is in the age group (25-34) and 0 otherwise. G3544 is a binary take the value 1 if the individual is in the age group (35-44) and 0 otherwise and the rest follow the same. Mstatus is a binary take the value 1 if the person is married and 0 otherwise. Training is a binary take 1 if the person had participated in training course and 0 otherwise. Secondary is a binary take 1 if the person has high school as the highest degree attained and 0 otherwise. University is a binary take 1 if the person has university degree as the highest degree attained and 0 otherwise. MAPhD is a binary take 1 if the person has high either masters or PhD as the highest degree attained and 0 otherwise.
Now I used the following command for obtaining the averages by gender and subsequently decompose the pooled sample. The margin command is as follow:
margins, over(Female) at(Female=(0 1))
My goal is to explain the gender gap in unemployment and investigated whether the gap is attributed to productive characteristics or potential discrimination against female.
I have seen the great work by Richard Williams as well as the decomposition in Stata Syntax, but the margin command can be used in many ways which got me a bit confused.
Could you please advise me whether the margin command that I used was correct or not? I used different “caned packages” such as (mvdcmp) and (fairlie) and got different results from each technique. I would greatly appreciate your help in explaining why I am getting all these different results?
Thank you very much in advance
I used a probit command to run just simple probit regression. And after fitting the model I wanted to obtain the average probability by gender.
Let’s assume the following model fitted using probit technique:
probit Unemp Female G2534 G3544 G4554 G5564 Mstatus Training Secondary University MAPhD
where Unemp is a binary take 1 if the individual unemployed and 0 otherwise. Female is a binary take 1 if female and 0 otherwise. G2534 is a binary take the value 1 if the individual is in the age group (25-34) and 0 otherwise. G3544 is a binary take the value 1 if the individual is in the age group (35-44) and 0 otherwise and the rest follow the same. Mstatus is a binary take the value 1 if the person is married and 0 otherwise. Training is a binary take 1 if the person had participated in training course and 0 otherwise. Secondary is a binary take 1 if the person has high school as the highest degree attained and 0 otherwise. University is a binary take 1 if the person has university degree as the highest degree attained and 0 otherwise. MAPhD is a binary take 1 if the person has high either masters or PhD as the highest degree attained and 0 otherwise.
Now I used the following command for obtaining the averages by gender and subsequently decompose the pooled sample. The margin command is as follow:
margins, over(Female) at(Female=(0 1))
My goal is to explain the gender gap in unemployment and investigated whether the gap is attributed to productive characteristics or potential discrimination against female.
I have seen the great work by Richard Williams as well as the decomposition in Stata Syntax, but the margin command can be used in many ways which got me a bit confused.
Could you please advise me whether the margin command that I used was correct or not? I used different “caned packages” such as (mvdcmp) and (fairlie) and got different results from each technique. I would greatly appreciate your help in explaining why I am getting all these different results?
Thank you very much in advance
