hetprobit command

[Sus Bau]

Hi all,

What is the correct command for hetprobit in stata? If the dep. variable is y and the indep. variables are x1, x2, x3 and you assume that x3 is a variable on which the variance depends, then would you put the command as "hetprobit y x1 x2, het(x3)" or would it be "hetprobit y x1 x2 x3, het(x3)"?

I tried both. In the first case I found the result favouring heterosk. and in the latter case, no heterosk.

Your help is much appreciated.

Thank you,

Sus

[Marcos Almeida]

Hello Sus,

Welcome to the Stata Forum/Statalist.

According to the Stata Manual, the syntax for - hetprobit is:

hetprobit depvar [indepvars] [if] [in] [weight], het(varlist [, offset(varname_o)]) [options]

Being this so, you are supposed to include in the het() option the "het" variable(s).

In the example 1 of the Stata Manual, on page 1009, we see that "xhet" is a different variable, if compared to "x".

Hopefully that helps.

[Richard Williams]

It isn't unusual to have the same variable in both equations. But the more overlap there is between the two equations, the greater the problem of multicollinearity may be. As the manual entry for hetprobit states,

As stated elsewhere (Greene 2012, 714), this is a difficult model to fit, and it is not uncommon for it to require many iterations or for the optimizer to print out warnings and informative messages during the optimization. Slow convergence is especially common for models in which one or more of the independent variables appear in both the index and variance functions.

[Sus Bau]

Many thanks Marcos and Richard for your help.