Statalist The Stata Listserver

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

st: ivreg2 - some basic clarifications please

From   Kit Baum <>
Subject   st: ivreg2 - some basic clarifications please
Date   Fri, 13 Oct 2006 10:12:50 -0400

Ramani said

I an very new to instrumenting and have run fairly simple regressions
with cross-section data before. I need to instrument a variable and
would be grateful for some clarifications about things that may seem
very obvious to you.
1. Is ivreg2 only for use with panel or time series data? Can I use it
to instrument a regression with cross-section data? It is not quite
clear in the help file.
2. Is the Sargan test reported only relevant for panel data analysis?
Can it be used for testing over-identification in cross-section
3. What is the test of excluded instruments that ivreg2 reports? Is it
the Stock and Yogo test?
I tried instrumenting with ivreg and it didn't report an R-squared for
the 2-SLS regression. Can't figure out why. So I thought I would try
ivreg2, if possible.

I am one of the authors of ivreg2.

1) ivreg2 is first of all an extended version of ivreg, which fits IV regressions in any context. It is often used on cross-sectional data, but can also be used for pure time-series data, pooled CS-TS data, or panel data. If you look at the clickable examples in ivreg2 help you will see a number of cross-sectional regressions at the outset.

2) The Sargan (or, if GMM is used, Hansen) test is a test which may be applied to any IV model which is overidentified.

3) There is a test for excluded instruments, described at some length in the help file, which is the Anderson-Rubin statistic, available in the first stage output. Stock and Yogo also recommend the Cragg- Donald F statistic, as the help file indicates. The help file is admittedly long and tedious, but reading through it (and the cited references) may be quite helpful.


Kit Baum, Boston College Economics
An Introduction to Modern Econometrics Using Stata:

* For searches and help try:

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