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

st: newey2 and first-stage results

From   "Todd D. Kendall" <>
Subject   st: newey2 and first-stage results
Date   Wed, 25 Jun 2008 16:59:44 -0500

I have learned a lot from reading Statlist -- thanks to everyone who
writes and answers questions.  Now I have a question of my own, and I
haven't been able to find anything in the archives so far.

I am trying to figure out just what newey2 does with instruments.
When I specify

. newey2 y (x1 =z1) x2, lag(1) first

STATA gives me just the first-stage results I would expect -- in fact,
I can replicate these first stage results with

. reg x1 z1 x2

However, if I try taking the predicted values from this first stage:

. predict x1hat

and "manually" running the IV results:

. newey y x1hat x2, lag(1)

I get the right coefficients (obviously), but the wrong standard
errors.  I surmise that newey2 must be doing something special with
the part of the covariance matrix associated with the first-stage
fitted values.  I have tried to read through the newey2 ado file, but
can't see exactly what's going on.

I'd appreciate any thoughts from anyone who understands newey2 better than I.

*   For searches and help try:

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