www.cmm.bristol.ac.uk/forum

Hello, forum,

I have two questions.

I'm running mlwin from within Stata and am using the ivmlwin macro to estimate a three-level growth model (age nested within kids nested within mothers). I would like to get the first stage F from the two-stage model. This was not doable about a year ago. Can it be done now?

If not, may I get it by running a logistic regression on the variable I want to instrument for?

Second question: Can I do robustness checks on my IV models using runmlwin?

Thanks,
Diego

Hi Diego,

I'm afraid we have not looked further at this since we last spoke. Basically there have never been any automatic IV routines in MLwiN. The only work we are aware of is the user-written ivmlwin macro which you refer to. One might be able to manually do all types of IV analyses in MLwiN (and therefore runmlwin), but this is not an area where I have particular expertise so cannot advise.

I believe an alternative to IV is to fit simultaneous equation models, so you could explore this approach. In your case multilevel simultaneous equation models. You might therefore want to take a look at the aML software or the -gsem- command in Stata which can fit multilevel simultaneous equation models with responses defined at different levels of analysis.

Good luck

George