Hi,

I am in the process of trying to run a Bivariate Random Effects Probit (Plum, 2016) to analyse the probability of married couples retiring based on Australian superannuation.

I am curious as how marginal effects should be calculated.

Our estimation code is:

Code:
/. bireprob retired_1 age ///
>         (retired_2 age_2)
and then esitmating margins for age results in

Code:
 margins, dydx(age age_2)

Average marginal effects                        Number of obs     =      2,836
Model VCE    : OIM

Expression   : Linear prediction, predict()
dy/dx w.r.t. : age age_2

------------------------------------------------------------------------------
             |            Delta-method
             |      dy/dx   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         age |   .1838857   .0119246    15.42   0.000      .160514    .2072574
       age_2 |          0  (omitted)
------------------------------------------------------------------------------

The fact that age_2 is omitted seems that either the STATA command margins doesn't account for the nature of the estimation, or I am making a mess of it.

Any thoughts would be appreciated.


refs
https://ideas.repec.org/a/tsj/stataj...i1p96-111.html

edit*
Unfortunately, due to the confidentiality requirements we can't link the data for replication, but it is sourced from HILDA ( Household, Income and Labour Dynamics in Australia, https://melbourneinstitute.unimelb.edu.au/hilda).