-------------------------------------------------------------------------------
help for meoprobit                                             (version 1.0.1)
-------------------------------------------------------------------------------

Compute marginal effects at means an their standard errors after oprobit

meoprobit [in] [if] [, nodiscrete stats(statistics) variance exp]

where statistics can be one of the elements sd, z or p.

Description

meoprobit computes marginal effects at means and their standard errors in an ordered probit model. meoprobit is executed after Stata's oprobit command.

meoprobit computes the marginal effects and standard errors (based on the delta method) analytically and is therefore faster than mfx and -mfx2-, which proceed numerically.

An alternative to meoprobit is -margeff-, which has the advantage of greater generality, more options and is better linked with other Stata commands such as lincom and others after estimation.

The advantages of meoprobit are its speed, its compact way of displaying the results and the fact that it is capable to compute also the marginal effect of the regressors on the expected mean of the outcome variable.

(Note of caution: If we have chosen to estimate an ordered probit model then this is because we think that the dependent variable carries only ordinal information but that the scaling of the variable, apart from preserving the order, carries no meaning. Therefore, strictly speaking, computing and interpreting the mean value of the ordinal dependent variable makes no sense. The user should question whether in the specific context it seems justified to look at the mean of the dependent variable.)

Options

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

stats() specifying sd, z or p in parentheses determines whether standard errors (the default), z-values or p-values are displayed.

if, in These options serve to restrict the means and frequency computation to a sub sample and allow to compute marginal effects at the mean values for a specific subgroup. The underlying regression coefficients and standard errors remain those from the original estimation.

variance specifies that you want the covariance matrix of the marginal effects to be saved as matrix V.

exp requests the computation of the marginal effect of the regressors on the expected mean of the outcome variable.

Output

The columns labeled c1, c2 ... cn refer to the categories of the dependent variable. If the option exp is chosen, the column labeled E[Y] refers to the expected mean of the dependent variable.

Saved results

PS The row vector PS contains sample frequencies for the categories of the dependent variable and its sample mean.

P The row vector P contains probabilities predicted at means for the categories of the dependent variable and the resulting predicted sample mean.

RES The matrix RES contains the results: marginal effects and standard errors (or z-values or P-values).

V If the options variance or exp are chosen, the matrix V contains the covariance matrix of the marginal effects.

Example

. oprobit y x

. meoprobit, stats(p)

Author

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

Also see