help probitmiss -------------------------------------------------------------------------------

Title

Efficient Estimator for Probit model with missing data

Syntax probitmiss [depvar] [varlistw] [varlistx] [,] [options]

options Description -------------------------------------------------------------------------

numw specifies the number of variables with missing data ñ default is numw(1)

-------------------------------------------------------------------------

Description

probitmiss provides efficient parameter estimates for a probit model of [ depvar] on the variables listed in [varlistw] and [varlistx] when the variables listed in [varlistw] have missing values. The command is implemented by typing probitmiss followed by [depvar] [varlistw] [varlistx]. [varlistw] is the set of explanatory variables for which some observations are missing and [varlistx] is the set of explanatory variables with no missing data. The conditions needed for efficiency are discussed in more detail in Conniffe and OíNeill (2011)

probitmiss also reports the results of a Hausman type test for the Missing at Random assumption of the estimator.

Example 1

--------------------------------------------------------------------------- . probitmiss riskydum2 rho_new eta eta2 college woman staciv_1 area5_2 area5_3 area5_4 area5_5, numw(1)

This estimates a probit model of riskydum2 on rho_new-area5_5 when only rho_new contains missing data.

---------------------------------------------------------------------------

Example 2

--------------------------------------------------------------------------- . probitmiss riskydum2 rho_new eta college woman staciv_1 area5_2 area5_3 area5_4 area5_5, numw(3)

This estimates a probit model of riskydum2 on rho_new-area5_5 when rho_new, eta and college all contain missing data.

---------------------------------------------------------------------------

Saved Results

probitmiss saves the following in e():

Scalars e(N) number of complete observations e(N2) total sample size(including incomplete observations) e(Chi2) Chi-squared statistic for Hausman test of MAR

Matrices e(b) coefficient vector e(V) variance-covariance matrix of the estimator

References

Conniffe,D. and D.OíNeill (2011) Efficient Probit Estimation with Partially Missing Covariates,î forthcoming in Advances in Econometrics: Volume 27, Missing Data Methods editor D. Drukker.