[Andrew Musau]

The estimator will implement listwise deletion, so in terms of the estimation itself, missing values are not a problem. Whether your results are biased due to the missing values depends on the mechanism behind the missingness. If these observations are random missing, then you are fine with listwise deletion. This is not the case with non-random missing values.

[Anne Wanyonyi]

After running a simple heckprobit model, I get this error message:

 heckprobit minimumwage educ age, select(married children educ age)
selection equation:
outcome does not vary; remember:
                                  0 = negative outcome,
        all other nonmissing values = positive outcome

this is how minimumwage looks like:

 type:  numeric (float)

                 range:  [0,1]                        units:  1
         unique values:  2                        missing .:  0/2,000

            tabulation:  Freq.  Value
                           395  0
                         1,605  1

How can I solve this?

[Anne Wanyonyi]

Going back to gsem, I think I will be able to run the model

 gsem (minimumwage <- educ age L, logit)(selected <- married children educ age L@1,family(gaussian, udepvar(notselected))), var(L@1 e.minimumwage@a e.selected@a)

once I figure out what

depvar1 <- ... , family(gauss, udepvar(depvar2)

should be when the family is bernoulli and link is logit/probit.

This is because I think gaussian is used when the response variable is continuous.

I haven't come across a suggestion in the manual, but I will appreciate any suggestions of how I can address that.

[Andrew Musau]

After running a simple heckprobit model, I get this error message:
Code:

heckprobit minimumwage educ age, select(married children educ age)

selection equation:
outcome does not vary; remember:
0 = negative outcome,
all other nonmissing values = positive outcome

You have not specified an outcome in your selection equation. The syntax is

 heckprobit depvar indepvars [if] [in] [weight] , select([depvar_s =] varlist_s [, noconstant offset(varname_o)]) [options]

where the outcome in the selection equation precedes the equals sign.

#18: I happen to think that the gsem syntax that replicates heckprobit is much more complicated than you have been attempting. A good place to start trying to figure this out is David Drukker's presentation from the 2014 German Users Meeting.

http://fmwww.bc.edu/RePEc/dsug2014/d...ukker_gsem.pdf