Cheking on Methodology

[Zino Faretti]

Hello, I am using Stata 17 and I am doing a research using a probit model. My data is a firm level data and cross-sectional belonging to 28 countries. I did include the i.country in my regressions to ensure country dummy variables. My data set has 2 sectors (agriculture and services) with total observations of 8000. Is it okay to do two regressions to see the impact of the independent variable on the dependent variable using the "if" statement. (When used the agricultural sector observations drop to 3500 and those of the services sector to 4500).

I am doing so in order to move from the general sample to a more specific analysis (sectoral)

Is any misconception in this approach?

Thank you for your support

[George Ford]

Sure, you can do it. If you think the effect of the X's are different.

You can also use interactions to get differences in coefficients, or have dummies for each group and estimate separate coefficients for each group in a single model.

Interpreting interaction terms in non-linear models is a bit tricky, but I think there's something in ssc that does it for you.

[Zino Faretti]

Thank you so much Mr. Ford for your kind reply!