Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: exclusion restriction

From   Austin Nichols <>
Subject   Re: st: exclusion restriction
Date   Sat, 10 Apr 2010 14:56:28 -0400

Paula <>:
Yes, you have to tell a compelling story about why we should believe
the exclusion restriction, as it is not directly testable.  You should
also report the diagnostic output from a near-equivalent -ivreg2-
specification that corresponds to the linear probability model.

On Sat, Apr 10, 2010 at 10:34 AM,  <> wrote:
> Dear Statalist,
> I have estimated a bivariate probit using stata biprobit in order to analyse
> the effect of care by a grandparent (GP) on the mother's probability to
> work. However, I have not included in the GP equation a variable that is
> related to GP care but uncorrelated with the error term in the outcome
> equation, as would be advisable. I do not have many variables to choose from
> and the only one that seems to come close is the fact that the father of the
> mother (a grandfather, who is not necessarily the carer) is alive.
> My doubts are:
> I)                    If this variable  is  related to GP care but
> uncorrelated with the error term in the outcome equation, do I have to
> explain the logic of the relation, or is it sufficient that there is a
> relation?
> 2)           What tests should I make to assess whether it is a good choice?
> The tests that I've seen suggested are obtainable after IVreg, not after
> biprobit, I think.
> I am using Stata version 9.2.

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index