I am doing Chow test to compare coefficients of two separate groups of samples. Such as male/ female, migrant/native. My regression model is a Mincer-type equation. The dependent variable is log wage. Most of my independent variables are factor/ categorical variables. For Chow test at first I generated groups
Code:
gen group1= frgn == 1
gen group2=group1==0
Then regression equations are
Code:
reg lhrw i.educ exper expr2 sex i.prvdmy han maris if group1 == 1
reg lhrw i.educ exper expr2 sex i.prvdmy han maris if group2 == 1
Then I did
Code:
gen edu1 = educ*group1
gen exper1 = exper*group1
gen expsq1 = expr2*group1
gen sex1 = sex*group1
Also same for group 2. I was not able to put i.educ although it is a factor variable with 4 categories.
Pooled regression equation
Code:
reg lhrw group1 edu1 exper1 expsq1 sex1 prvdmy1 han1 maris1 group2 edu2 exper2 expsq2 sex2 prvdmy2 han2 maris2, noconstant
Then I did the CHOW test
Code:
test _b[edu1]=_b[edu2], notest
....................................
test _b[group1]=_b[group2], accum
Am I doing ok as I was not able to write i.educ or i.prvdmy? I already got a test result that we can reject the null that means the difference is not zero. Significant difference.
0 Response to Chow test where most controls are factor/ categorical variables
Post a Comment