Hi,

I have two treatment variables, a and b, both are dummies. Then I
Code:
egen treatment=group(a b)
I run two identical -xtlogit- regression for two subsamples separated by variable a:
say the dependent variable is y, independent variable of interest is x, and set of controls C

Code:
xtlogit y x b C if a == 1 // x is positive at 1% significance level
Code:
xtlogit y x b C if a == 0 // x is positive at 1% significance level
Then i run the regression for the whole sample, I expect x to be positive and significant based on the results from subsample.
Code:
xtlogit y x i.treatment C //x is negative and insignificant.
What could be the potential reason of getting different signs?

Thanks a lot
Jasmine