Hi all,
Just to be clear, I'm really beginning with Stata and Econometrics in general, so my questions might sound a bit stupide. Anyway, here's what I'm trying to do for a project:
I am interested in what determinants may drive a car manufacturing firm to sell its car only in specific countries. The same car can obviously be available on different markets, but from one market to another, the choices might be strongly correlated.
That's why I'm going for a multivariate probit. Let's say for now that we have three markets: US, GB and Germany. The choice to sell a car on a specific market might depend on the car itself (for example the mass of the vehicle as a proxy of the type), on the firm characteristics (for example the size of the firm or the location of its head office) and on the market itself (for example the total number of cars in this country). What I'm doing for now is:
cmp ///
(US = car_mass firm_size location_US location_GB location_DE) ///
(DE = car_mass firm_size location_US location_GB location_DE) ///
(GB = car_mass firm_size location_US location_GB location_DE) ///
, ind(4 4 4) ghkdraws(10)
or
mvprobit ///
(US = car_mass firm_size location_US location_GB location_DE) ///
(DE = car_mass firm_size location_US location_GB location_DE) ///
(GB = car_mass firm_size location_US location_GB location_DE) ///
, draws(10)
where US/DE/GB indicates if the car is available for sell in the country, car_mass and firm_size are continuous variables and location_XX are dummies (head office in country XX). Everything works well. I know I probably need to increase the draws as it is a simulated method...
1) My problem is when I want to add a country-level variable, such as the number of cars in the country for example. But that's just one scalar for each probit equation. That might not be relevant for 3 countries only, but maybe for 20 different countries. How can I do that in this framework and how would that translate in Stata?
2) Some cars are from the same manufacturer. Can I add random effects for the firms? How should I do that?
3) Is cmp/mvprobit better than the other and what about multivariate logit that seems less usual from what I've seen?
Thanks for the help,
D.
Just to be clear, I'm really beginning with Stata and Econometrics in general, so my questions might sound a bit stupide. Anyway, here's what I'm trying to do for a project:
I am interested in what determinants may drive a car manufacturing firm to sell its car only in specific countries. The same car can obviously be available on different markets, but from one market to another, the choices might be strongly correlated.
That's why I'm going for a multivariate probit. Let's say for now that we have three markets: US, GB and Germany. The choice to sell a car on a specific market might depend on the car itself (for example the mass of the vehicle as a proxy of the type), on the firm characteristics (for example the size of the firm or the location of its head office) and on the market itself (for example the total number of cars in this country). What I'm doing for now is:
cmp ///
(US = car_mass firm_size location_US location_GB location_DE) ///
(DE = car_mass firm_size location_US location_GB location_DE) ///
(GB = car_mass firm_size location_US location_GB location_DE) ///
, ind(4 4 4) ghkdraws(10)
or
mvprobit ///
(US = car_mass firm_size location_US location_GB location_DE) ///
(DE = car_mass firm_size location_US location_GB location_DE) ///
(GB = car_mass firm_size location_US location_GB location_DE) ///
, draws(10)
where US/DE/GB indicates if the car is available for sell in the country, car_mass and firm_size are continuous variables and location_XX are dummies (head office in country XX). Everything works well. I know I probably need to increase the draws as it is a simulated method...
1) My problem is when I want to add a country-level variable, such as the number of cars in the country for example. But that's just one scalar for each probit equation. That might not be relevant for 3 countries only, but maybe for 20 different countries. How can I do that in this framework and how would that translate in Stata?
2) Some cars are from the same manufacturer. Can I add random effects for the firms? How should I do that?
3) Is cmp/mvprobit better than the other and what about multivariate logit that seems less usual from what I've seen?
Thanks for the help,
D.
