```-------------------------------------------------------------------------------
help for mehetprob                                             (version 1.0.0)
-------------------------------------------------------------------------------

Compute marginal effects at means an their standard errors after hetprob

mehetprob [, nodiscrete]

Description

mehetprob computes marginal effects at means and their standard errors in a
heteroscedastic probit model.  mehetprob is executed after Stata's hetprob
command. It uses the coefficients and standard errors returned by hetprob in
order to compute marginal effects at means and their standard errors by using
the delta method. The syntax is such that it requires at least Stata 8.

mehetprob delivers the same results as

. mfx compute

after hetprob.

While mfx computes the derivatives needed to infer the standard errors
numerically, mehetprob computes them analytically and is therefore considerably
faster. It takes only a few seconds while mfx compute after hetprob can take
several hours in Stata 8 and several minutes in Stata 9.

Options

nodiscrete specifies that dummy (indicator) variables be treated as continuous.

Saved results

XBAR        The matrix XBAR contains means of the regressors of the main
>  equation of the
heteroscedastic probit model.

ZBAR        The matrix ZBAR contains means of the regressors of the equa
> tion to model the
variance in the heteroscedastic probit model.

DER         The element (i,j) of the matrix DER contains the derivative
> of the marginal effect of the
regressor w_i with respect to the regression coefficient the
> ta_j.
The delta method is based on this matrix.

C           The matrix C contains the results: marginal effects, standar
> d errors,
z-values and P-values derived therefrom.

mehetprob temporarily creates the matrices C, C1, C2, ZG, ZGx, XB, XBx,
> B, ZBAR, XBAR, G, D, V.
If any of this matrices exists prior to the execution of the command, it
>  will be overwritten.

Example

. hetprob x y, het(y z)

. mehetprob

Author

Thomas Cornelißen, University of Hannover, Germany
cornelissen@mbox.iqw.uni-hannover.de

Also see

```