[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

# st: Date: Mon, 21 Aug 2006 17:35:17 +0100

 From "Jennifer Leavy" To Subject st: Date: Mon, 21 Aug 2006 17:35:17 +0100 Date Mon, 21 Aug 2006 12:35:50 -0400 (EDT)

```Mark, thanks so much, that's very helpful.
i'll try what you suggest and see if it tells me what I want to know
Jennifer

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of
> Jennifer Leavy
> Sent: Monday, August 21, 2006 2:25 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: st: RE: how to cluster with ivprobit with two-step option?
>
> Dear Mark
>
> I am trying to model the marketing decisions of smallholder
> farmers, not all of whom sell their crops. So I start with a
> heckman selection model with a probit in the first stage,
> i.e. dependent variable is whether or not that person sells
> their crops, and in the second stage the outcome equation has
> as the dependent variable how much is sold, given that the
> person does actually sell. So that takes care of the sample
> selection bias. There is also potential endogeneity in that
> some of the regressors are likely to be endogenous. So i need
> to incorporate instrumental variables in both the selection
> ('first stage') and outcome ('second stage') equations. So I
> am referring to stages in the heckman selection model. In
> fact, as you say, each of those stages has two stages.
>
> I hope I make sense!
>
> But your comment about what the cluster option doessolves my
> problem, I think  - I can forget about using cluster in the
> first stage and use the two-step option rather than MLE, then
> take the inverse mills ratio from this for the second stage
> of the heckman?

The outline of your model is:

1a.  Eqn(s) for endogenous regressors in selection eqn.

1b.  Selection equation, explicitly modelled with endogenous regressors.

2a.  Eqn(s) for endogenous regressors in outcome eqn.

2b.  Outcome eqn suffers from selection bias (2a) and endogeneity bias
(1b).

If the equation of interest is the outcome equation of the selection
model, it isn't clear that you need to estimate explicitly the selection
equation explicitly, i.e., 1a and 1b.  In other words, you're talking in
terms of estimating a system of equations, but you may only need to
worry about just the outcome equation.  If that's the case, then, for
example, the excluded instruments you were thinking of using in (1a) and
(1b) to instrument for the endogenous regressors in the selection
equation could be used directly as the exclusion restrictions in the
Heckman-type estimation in (2b).  Or, put another way, the probit first
stage in the selection estimation can be a reduced form estimation with
just the exogenous regressors.

HTH.

Cheers,
Mark

Prof. Mark E. Schaffer
Director
Centre for Economic Reform and Transformation
Department of Economics
School of Management & Languages
Heriot-Watt University
Edinburgh EH14 4AS  UK
44-131-451-3494 direct
44-131-451-3296 fax
http://www.sml.hw.ac.uk/cert

>
> Jennifer
>
> > -----Original Message-----
> > From: owner-statalist@hsphsun2.harvard.edu
> > [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Jennifer
> > Leavy
> > Sent: Monday, August 21, 2006 11:00 AM
> > To: statalist@hsphsun2.harvard.edu
> > Subject: st: how to cluster with ivprobit with two-step option?
> >
> > Dear all
> >
> > I am trying to run a heckman selection model with potentially
> > endogenous regressors in both the selection and outcome
> equations. I
> > used ivprobit in the first stage, and need to use the
> cluster option
> > because my data are from a household survey and some of the
> > respondents come from the same household. This means I
> cannot use the
> > twostep option with ivprobit as cluster is not allowed. But
> as I have
> > more than one potentially endogenous regressor and
> therefore several
> > instruments
>
> Can you say more precisely what the structure of your model
> is?  I'm a little confused by using "ivprobit in the first
> stage".  ivprobit with the twostep option does itself have
> its own first stage.  Plus the cluster option doesn't change
> coefficient estimates, only the SEs.  If you are using
> ivprobit in a first stage, then you wouldn't need the SEs.
> ... Which is why I'm a little confused.
>
> --Mark
>
> Prof. Mark E. Schaffer
> Director
> Centre for Economic Reform and Transformation Department of
> Economics School of Management & Languages Heriot-Watt
> University Edinburgh EH14 4AS  UK
> 44-131-451-3494 direct
> 44-131-451-3296 fax
> http://www.sml.hw.ac.uk/cert
>
> > I find that the model is not converging with the MLE
> default - I have
> > been pressing break after a few thousand iterations. I've
> been unable
> > to find a user-written programme that allows the cluster
> option with
> > twostep. I would very much appreciate any suggestions as to
> how I can
> > get round this problem (I'm using STATA 9), for example, would I be
> > able to alter the code in the ivprob programme to allow me
> to do this?
> > Or is there a statistical reason why it's not already built
> in to the
> > ivprobit, twostep com  mand?
> >
> > Many thanks
> >
> > Jennifer
> >
> >
> >

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```

 © Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index