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

Re: st: RE: Std. -ivregress- vs. "by hand" using -reg- with nonlinear endogenous variable

From   Austin Nichols <[email protected]>
To   [email protected]
Subject   Re: st: RE: Std. -ivregress- vs. "by hand" using -reg- with nonlinear endogenous variable
Date   Thu, 7 Jan 2010 10:04:52 -0500

Misha Spisok--
I would guess that the bias is so small in your case that even OLS
will look consistent (you should always compare to OLS in these types
of examples).  The "forbidden regression" in which you square the
predicted value for y2 likely has an asymptotic bias smaller than OLS
but larger than a consistent IV strategy, though I am just guessing at
that.  Note that a single example that fails to demonstrate the
inconsistency of an estimator that is known not to be consistent in
many specific cases is not a proof that an estimator is consistent;
the problem in the Cameron&Trivedi book asks you to construct a case
that *does* demonstrate inconsistency.  Since Statalist has a
prohibition on homework problems, and my guess is that some people on
the list or browsing the archives will face this exact question on
their homework, I don't think a solution to the problem in the
Cameron&Trivedi book should be posted here...

On Wed, Jan 6, 2010 at 3:24 PM, DE SOUZA Eric
<[email protected]> wrote:
> Question 2: using OLS for the second stage does not take into account that you are using estimated variables instead of the original endogenous variables. For this reason, the correct standard errors are not those calculated under OLS
> Question 1: See slide 53 onwards in the presentation by Christopher Baum:
> Eric de Souza
> -----Original Message-----
> From: [email protected] [mailto:[email protected]] On Behalf Of Misha Spisok
> Sent: 06 January 2010 21:11
> To: [email protected]
> Subject: st: Std. -ivregress- vs. "by hand" using -reg- with nonlinear endogenous variable
> Hello, Statalist!
> [There have been no replies to a previous message posted nine days ago.]
> My questions, in short, are the following:
> 1. How is it "obvious" from my generated data and approach that one estimator is consistent while the other is inconsistent?
> 2.  Why do the standard errors differ when doing IV "by hand" (i.e., using -reg- for each stage) versus using -ivregress-?

*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index