Oaxaca Style Decomposition for Probit with Sample Selection

[Joelle Saad-Lessler]

I have a Probit model with Heckman sample selection. I would like to do a decomposition of changes in the dependent variable over two time periods, Oaxaca Blinder style. Nldecompose and Fairlie decompositions do not allow for sample selection.
Any other possibilities?

[George Ford]

I suppose you could generate the Inverse Mills Ratio from the probit equation and include it in the decomposition, but you'd need to adjust the covariance matrix to account for the prediction (Murphy Topel, or bootstrap).