Bootstrap with Oaxaca Model

[Katherine Oleas]

Good morning to everybody

I'm running an Oaxaca Blinder model and I want to include vce(bootstrap) but when I run the command it tells me that I can't use it with svy. Can somebody help me


svyset upm [iw=fexp], strata (estrato) vce(linearized) singleunit(certainty)
oaxaca lingreso escol exper exper2 jefe ecivil tamaño condicion normalize(b.provincia1 provincia2 provincia3 provincia4 provincia5 provincia6 provincia8 provincia9 provincia10 provincia11 provincia12 provincia13 provincia14 provincia15 provincia16 provincia17 provincia18 provincia19 provincia20 provincia21 provincia22 provincia23) normalize(b.raza1 raza2 raza3) normalize(b.actividad1 actividad2 actividad3 actividad4) mills, by(mujer) vce(bootstrap) omega noisily relax swap svy adjust(mills)

[FernandoRios]

This is just as it is. Bootstrap cannot be combined with weights.
Your options will be:
Option 1 Use asymptotic standard errors and not bootstrap Standard errors.
Option 2 Obtained bootstrap weights. But those need to be calibrated and are usually provided by the agency that collected the data
Option 3 Run bootstrap without weights
Option 4 Run bootstrap with a wrapper to allow for weights (but incorrectly)

[Katherine Oleas]

Another query I am running the Oaxaca model with the inverse ratio of mills. The coefficients should change a lot if I estimate without the inverse time of mills?

[FernandoRios]

They can
But Oaxaca Decomposition with Selection into groups is a hard question.
There is a review paper that shows you different ways of using the IMR and hw to interpret it.
see https://link.springer.com/article/10...28395.38694.4b