Use regression results from heckman twostep as initial values for heckman ml

[Zhongying Gan]

Hello, I'm trying to estimate a heckman model. The thing is when I use "heckman,twostep" it works, but when I use "heckman,ml", the maximum likelihood function won't converge. But I have to use "heckman,ml" because I need to use "margins, ycond" later and "ycond" option does not work after "heckman,twostep". That's why I want to try to put the estimates from the twostep estimation as the initial search values for the "heckman,ml" in the hope that it will converge this way.

. use https://www.stata-press.com/data/r17/womenwk,clear

. heckman wage educ age, select(married children educ age) twostep

Heckman selection model -- two-step estimates   Number of obs     =      2,000
(regression model with sample selection)              Selected    =      1,343
                                                      Nonselected =        657

                                                Wald chi2(2)      =     442.54
                                                Prob > chi2       =     0.0000

------------------------------------------------------------------------------
        wage |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
wage         |
   education |   .9825259   .0538821    18.23   0.000     .8769189    1.088133
         age |   .2118695   .0220511     9.61   0.000     .1686502    .2550888
       _cons |   .7340391   1.248331     0.59   0.557    -1.712645    3.180723
-------------+----------------------------------------------------------------
select       |
     married |   .4308575    .074208     5.81   0.000     .2854125    .5763025
    children |   .4473249   .0287417    15.56   0.000     .3909922    .5036576
   education |   .0583645   .0109742     5.32   0.000     .0368555    .0798735
         age |   .0347211   .0042293     8.21   0.000     .0264318    .0430105
       _cons |  -2.467365   .1925635   -12.81   0.000    -2.844782   -2.089948
-------------+----------------------------------------------------------------
/mills       |
      lambda |   4.001615   .6065388     6.60   0.000     2.812821     5.19041
-------------+----------------------------------------------------------------
         rho |    0.67284
       sigma |  5.9473529
------------------------------------------------------------------------------

To put in the initial search value from the above estimation results into "heckman, ml" estimation, I know it's something like

heckman wage educ age, select(married children educ age) ml from()

but I don't know what to put into the parentheses after "from".

from() specifies initial values for the coefficients. Not all estimators in Stata support this option. You can specify the initial values in one of three ways: by specifying the name of a vector containing the initial values (for example, from(b0), where b0 is a properly labeled vector); by specifying coefficient names with the values (for example, from(age=2.1 /sigma=7.4)); or by specifying a list of values (for example, from(2.1 7.4, copy)).

This is from https://www.stata.com/manuals/rmaximize.pdf.

I know I can type in the values of the estimates into from() one by one, but in my own model, I have more than 50 variables, and I just cannot type it one by one. I was wondering whether there is an easier way. Thank you.

[FernandoRios]

Hi Zhongying
What you ask is straight forward, however, it may or may not help with your problem trying to estimate heckman-ml model.
Here an example:

use https://www.stata-press.com/data/r17/womenwk,clear
heckman wage educ age, select(married children educ age) twostep
matrix b=e(b)
heckman wage educ age, select(married children educ age) ml from(b,skip)

HTH

[Zhongying Gan]

Hi Zhongying
What you ask is straight forward, however, it may or may not help with your problem trying to estimate heckman-ml model.
Here an example:

use https://www.stata-press.com/data/r17/womenwk,clear
heckman wage educ age, select(married children educ age) twostep
matrix b=e(b)
heckman wage educ age, select(married children educ age) ml from(b,skip)

HTH

Hello Fernando,

Thank you so much. It worked. Unfortunately, using the estimates from Heckman twostep didn't help with the convergence of estimating the model using MLE.

[FernandoRios]

That is what I imagine would happen.
Perhaps something else to consider is to modify your model specification. 50 variables seems a lot. And its possible that you have a few that are creating the problem
I would start analyzing the probit, and see which variables are problematic.
Based on the heckman two step, for example, I would start by considering excluding those with very large standard errors (if any)
HTH