vce(bootstrap) for Mixed Model

[Monica Muller]

Hi,

I need to bootsrap the standard error for one variable in my model. It's the inverse mills ratio that I calculated manually from the first stage of Heckman model and I am bringing it to the second stage model which is a repeated measurement model and I am using mixed model for it. I want to bootsrap the SE to mitigate the bias in SE caused by manually bringing in the IMR.

I want to run something like:

mixed y  `x' IMR year || id:, vce(bootstrap (_b[IMR]),rep(100) seed(12345))

but I get this error: vcetype 'bootstrap' not allowed

Do you know of a way to bootsrap SE in a mixed model?

Thanks

[Carlo Lazzaro]

Monica.
as per -mixed- entry, it seems that -bootstrap- is not supported.

[Rich Goldstein]

you can, however, bootstrap the model using the bootstrap command; see "help bootstrap"

[Filipe Rodrigues]

Hi everyone

I used Rich's suggestion and worked fine.

I was also wondering if it wouldn't be necessary to add an option to bootstrap to specify the clustering?
For instance, making use of Monica's initial model, should we add cluster(id) like below:

bootstrap, rep(100) seed(12345) cluster(id): mixed y `x' IMR year || id:,