#192 Hausman test for endogenous regressors

[Daehoon Nahm]

The Hausman test statistic produced when the tsls method is used seems to be different from what I can get from the formulas in the reference mentioned in User's Guide (either J times the F statistic or the Chi-square statistic). Could you please make the formula for the statistic available or direct me to the formula in the reference? Thank you.

[Sven S.]

Hi,
to be honest I couldn't find any precise reference in the docs (user
guide or command reference) for the TSLS command. So this may count as a
documentation bug per se, but OTOH I'm not sure what formulas you are
talking about. Please be more specific and/or provide an example.

I think for the panel case (classic RE vs. FE) at some stage gretl
switched from the standard Hausman statistic to another one based on an
auxiliary regression, but I'm not sure whether that change applies to
the TSLS case as well. (For certain reasons I think this change would
not be as necessary in the TSLS case compared to the panel case.)

Thanks,
Sven

Am 04.05.2015 um 05:36 schrieb Daehoon Nahm:

---

[bugs:#192] http://sourceforge.net/p/gretl/bugs/192 Hausman test for
endogenous regressors

Status: open
Group: v1.0_(example)
Created: Mon May 04, 2015 03:36 AM UTC by Daehoon Nahm
Last Updated: Mon May 04, 2015 03:36 AM UTC
Owner: nobody

The Hausman test statistic produced when the tsls method is used seems
to be different from what I can get from the formulas in the reference
mentioned in User's Guide (either J times the F statistic or the
Chi-square statistic). Could you please make the formula for the
statistic available or direct me to the formula in the reference? Thank you.

---

Sent from sourceforge.net because you indicated interest in
https://sourceforge.net/p/gretl/bugs/192/
https://sourceforge.net/p/gretl/bugs/192

To unsubscribe from further messages, please visit
https://sourceforge.net/auth/subscriptions/
https://sourceforge.net/auth/subscriptions

[Riccardo "Jack" Lucchetti]

I think that what the user has in mind is the reference to Davidson and MacKinnon (2004) in the help to the tsls command.

If I understand your points correctly, the small differences you get are due to the fact that in ols we use df-corrected hypothesis test, that are avoided in other commands such as tsls. A small script follows, to illustrate how you can replicate the Hausman test given by tsls exactly via ols:

open mroz87.gdt --quiet
list X = const WA
list Z = WMED WFED

# 1st stage
ols WE X Z --quiet
series u = $uhat

# 2nd stage
ols WW X WE u --quiet --no-df-corr
omit u --chi-square --quiet

# tsls
tsls WW X WE ; X Z

[Daehoon Nahm]

Dear Professor Lucchetti,

Thank you so much for your prompt and clear explanation.
Thank you also for making such a good software program available to
everyone. I have been using gretl for some of my courses. I like it and
students also love it as it is so user friendly. I hit the wall when I
tried to explain how the statistic was computed. I know exactly how it's
done now.

Kind regards,
Daehoon

On Mon, May 4, 2015 at 8:59 PM, Riccardo jacklucchetti@users.sf.net wrote:

I think that what the user has in mind is the reference to Davidson and
MacKinnon (2004) in the help to the tsls command.

If I understand your points correctly, the small differences you get are
due to the fact that in ols we use df-corrected hypothesis test, that are
avoided in other commands such as tsls. A small script follows, to
illustrate how you can replicate the Hausman test given by tsls exactly
via ols:

open mroz87.gdt --quietlist X = const WAlist Z = WMED WFED

1st stageols WE X Z --quietseries u = $uhat

2nd stageols WW X WE u --quiet --no-df-corromit u --chi-square --quiet

tslstsls WW X WE ; X Z

---

• [bugs:#192] http://sourceforge.net/p/gretl/bugs/192 Hausman test for
  endogenous regressors*

Status: open
Group: v1.0_(example)
Created: Mon May 04, 2015 03:36 AM UTC by Daehoon Nahm
Last Updated: Mon May 04, 2015 03:36 AM UTC
Owner: nobody

The Hausman test statistic produced when the tsls method is used seems to
be different from what I can get from the formulas in the reference
mentioned in User's Guide (either J times the F statistic or the Chi-square
statistic). Could you please make the formula for the statistic available
or direct me to the formula in the reference? Thank you.

---

Sent from sourceforge.net because you indicated interest in
https://sourceforge.net/p/gretl/bugs/192/

To unsubscribe from further messages, please visit
https://sourceforge.net/auth/subscriptions/

--

Daehoon Nahm

Department of Economics | Room 417, E4A
Macquarie University, NSW 2109, Australia

T: +61 2 9850 9615
http://mq.edu.au

[image: Macquarie University] http://mq.edu.au

CRICOS Provider Number 00002J. Think before you print.
Please consider the environment before printing this email.

This message is intended for the addressee named and may
contain confidential information. If you are not the intended
recipient, please delete it and notify the sender. Views expressed
in this message are those of the individual sender, and are not
necessarily the views of Macquarie University.