Bookmark and Share

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

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

Re: st: St: interpret the result of Hausman test

From   John Antonakis <John.Antonakis@unil.ch>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: St: interpret the result of Hausman test
Date   Mon, 23 Apr 2012 10:50:45 +0200
Yes. As per the message below, I would sooner trust the iv estimator (IF you have strong instruments). 
Best,
J.

__________________________________________

Prof. John Antonakis
Faculty of Business and Economics
Department of Organizational Behavior
University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny
Switzerland
Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305
http://www.hec.unil.ch/people/jantonakis

Associate Editor
The Leadership Quarterly
__________________________________________


On 23.04.2012 04:37, Hoang Dinh Quoc wrote:

Dear Prof.

Thank you for your help.

Yes, I am sure that I am using the same control variables in the models.

For reg: the syntax I used is:
.regress depvar indepvar1 indepvar2 indepvar3 indepvar4 endovar

For ivreg2:
.ivreg2 depvar indepvar1 indepvar2 indepvar3 indepvar4 (endovar = IV),
endog(endovar)

With this result, I think I can conclude that I have endogeneity problem,
right? So what to do in order to solve this problem?


Best,
Quoc

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of John Antonakis
Sent: Friday, April 20, 2012 5:21 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: St: interpret the result of Hausman test

Odd that your OLS estimates is not significant and the iv estimate is.
Perhaps others can shed light on this.

Are you sure you are including the same control variables (exogenous) in
each model?

What, precisely, is the syntax for the reg and ivreg2 models?

J.

__________________________________________

Prof. John Antonakis
Faculty of Business and Economics
Department of Organizational Behavior
University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny
Switzerland
Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305
http://www.hec.unil.ch/people/jantonakis

Associate Editor
The Leadership Quarterly
__________________________________________


On 20.04.2012 11:37, Hoang Dinh Quoc wrote:

Thank you very much for your explanation, Prof.

Yes, it seems to be quite different between iv and ols; for the variable x
(suspect var for endogenous), the model ols shows the coefficient is

.03589

and the p-value 0.615; but the ivreg2 shows coefficient .3302337 and p

value

0.020.
Did you mean that I would better take the ovreg2 for the final result?

Best,
Quoc


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of John Antonakis
Sent: Friday, April 20, 2012 3:53 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: St: interpret the result of Hausman test

According to the endog test, your regressor is probably endogenous
(given that you are close to the commonly-determined critical value of p
<   .05) and thus requires instrumenting.  Are the estimates of iv and ols
very different? If they are, and if your instruments are strong , which
they seem to be judging form the Anderson test and the Stock-Yogo
critical values, you may be better off trusting the inefficient iv
estimate, than the efficient (but probably inconsistent) OLS estimate.

See: http://www.stata.com/statalist/archive/2012-03/msg01264.html

Best,
J.

__________________________________________

Prof. John Antonakis
Faculty of Business and Economics
Department of Organizational Behavior
University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny
Switzerland
Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305
http://www.hec.unil.ch/people/jantonakis

Associate Editor
The Leadership Quarterly
__________________________________________


On 20.04.2012 10:18, Hoang Dinh Quoc wrote:

Thanks. Below is what I got by ivreg2 y (x = z), endog(x). You talked

about

the p-value 0.0600, right? Does this mean that we can conclude no
endogeneity problem?

Best,
Quoc



Underidentification test (Anderson canon. corr. LM statistic):
49.520
                                                      Chi-sq(1) P-val =

0.0000


----------------------------------------------------------------------------

--
Weak identification test (Cragg-Donald Wald F statistic):

53.345

Stock-Yogo weak ID test critical values: 10% maximal IV size

16.38

                                            15% maximal IV size

8.96

                                            20% maximal IV size

6.66

                                            25% maximal IV size

5.53

Source: Stock-Yogo (2005).  Reproduced by permission.


----------------------------------------------------------------------------

Sargan statistic (overidentification test of all instruments):

0.000

                                                    (equation exactly
dentified)
-endog- option:
Endogeneity test of endogenous regressors:

3.538

                                                      Chi-sq(1) P-val =

0.0600

Regressors tested:    sc_tie_weak


----------------------------------------------------------------------------

--
Instrumented:         sc_tie_weak
Included instruments: sc_tie_strong sex income_cat_07 alter_SIOPs

head_siops

                         market_close ethnic headage leader hhknown

access_cre

                         Cre_Con mass_media Road_constraint red_gre
                         no_extension_contact _Idistrict_2 _Idistrict_3
                         _Idistrict_4 _Idistrict_5 _Idistrict_6

_Idistrict_7

Excluded instruments: loan_bank_job




-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of John Antonakis
Sent: Friday, April 20, 2012 3:03 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: St: interpret the result of Hausman test

No. I meant -endog- and not -orthog-.

Do you have the latest version of ivreg2?

. which ivreg2
c:\ado\plus\i\ivreg2.ado
*! ivreg2 3.1.04  19mar2012
*! authors cfb&    mes
*! see end of file for version comments

If not, updated your ivreg2 file:

ssc install ivreg2, replace

Then redo the iv-regression and see what you get.

Best,
J.

__________________________________________

Prof. John Antonakis
Faculty of Business and Economics
Department of Organizational Behavior
University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny
Switzerland
Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305
http://www.hec.unil.ch/people/jantonakis

Associate Editor
The Leadership Quarterly
__________________________________________


On 20.04.2012 09:50, Hoang Dinh Quoc wrote:

Dear Prof. Antonakis,

Thank you very much for your suggestion.



For your suggestion:

hausman one two, sigmamore
What does that give?

The result is below; I guess something went wrong with this result,

right?

        b = consistent under Ho and Ha; obtained from regress

            B = inconsistent under Ha, efficient under Ho; obtained from
ivregress



        Test:  Ho:  difference in coefficients not systematic



                      chi2(1) = (b-B)'[(V_b-V_B)^(-1)](b-B)

                              =    -3.33    chi2<0 ==>     model fitted on

these

                                            data fails to meet the

asymptotic

                                            assumptions of the Hausman

test;

                                            see suest for a generalized

test

Your comment: "ivreg2 y (x = z), endog(x)". I guess you meant option
'orthog' right? Because endog did not work on my Stata; I am using Stata

10.

Below is the result; according to this result, as the P-value (0.0600)

is

bigger than 0.5, I guess I can conclude x is not endogenous, right?



----------------------------------------------------------------------------

--
Sargan statistic (Lagrange multiplier test of excluded instruments):
3.538
                                                       Chi-sq(1) P-val =
0.0600
-orthog- option:
Sargan statistic (eqn. excluding suspect orthogonality conditions):
0.000
                                                       Chi-sq(0) P-val =
.
C statistic (exogeneity/orthogonality of suspect instruments):
3.538
                                                       Chi-sq(1) P-val =
0.0600





Best,

Quoc







-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of John

Antonakis

Sent: Thursday, April 19, 2012 8:42 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: St: interpret the result of Hausman test



Do:



hausman one two, sigmamore



What does that give? If the hausman test is still NPD try:



ivreg2 y (x = z), endog(x)



Also, did you try it in sem as I suggested?



If the p value of the endogeneity test is<     .05 then x is endogenous.



However, if your sample is small the test might not have much power (so

I would be worried about endogeneity if<     .10). If you have good

reason

to believe that x is endogenous then the iv estimator should be

retained.


HTH,

J.



__________________________________________



Prof. John Antonakis

Faculty of Business and Economics

Department of Organizational Behavior

University of Lausanne

Internef #618

CH-1015 Lausanne-Dorigny

Switzerland

Tel ++41 (0)21 692-3438

Fax ++41 (0)21 692-3305

http://www.hec.unil.ch/people/jantonakis



Associate Editor

The Leadership Quarterly

__________________________________________





On 19.04.2012 10:39, Hoang Dinh Quoc wrote:


Dear Prof. Antonakis,
Thank you very much for your quick support.
I followed your suggestion:
"reg y x
est store one
ivregress 2sls y (x=z)
est store two
hausman one two"
And I got this result:
Test:  Ho:  difference in coefficients not systematic
                       chi2(1) = (b-B)'[(V_b-V_B)^(-1)](b-B)
                               =        3.31
                     Prob>chi2 =      0.0687
                     (V_b-V_B is not positive definite)
With is result, can I conclude that no endogeneity problem?
Thanks,
Best,
Hoang Dinh Quoc
-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of John

Antonakis

Sent: Thursday, April 19, 2012 3:23 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: St: interpret the result of Hausman test
Hi:
I am not quite sure what you have done here.
If you want to do this "by hand" do an augmented regression:
http://www.stata.com/support/faqs/stat/endogeneity.html
Else, use the -endog- option in the user-written program, ivreg2,
available from ssc (i.e., ssc install ivreg2, replace), e.g. (for
dependent variable y, endogenous regressor x, and instrument z):
ivreg2 y (x = z), endog(x).
Or do the usual hausman test via Stata, e.g.,
reg y x
est store one
ivregress 2sls y (x=z)
est store two
hausman one two
Finally, you can do this in the new Stata command, -sem- using maximum
likelihood:
sem (y<-x) (x<-z), cov(e.y*e.x)
The test of the correlation between the disturbances is the Hausman
test, as we explain in detail here:
Antonakis, J., Bendahan, S., Jacquart, P.,&      Lalive, R. (2010). On
making causal claims: A review and recommendations. The Leadership
Quarterly, 21(6). 1086-1120.
http://www.hec.unil.ch/jantonakis/Causal_Claims.pdf
For more basic explanations see:
Antonakis, J., Bendahan, S., Jacquart, P.,&      Lalive, R. (submitted).
Causality and endogeneity: Problems and solutions. In D.V. Day (Ed.),
The Oxford Handbook of Leadership and Organizations.
http://www.hec.unil.ch/jantonakis/Causality_and_endogeneity_final.pdf
HTH,
J.
__________________________________________
Prof. John Antonakis
Faculty of Business and Economics
Department of Organizational Behavior
University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny
Switzerland
Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305
http://www.hec.unil.ch/people/jantonakis
Associate Editor
The Leadership Quarterly
__________________________________________
On 19.04.2012 10:14, Hoang Dinh Quoc wrote:
      >      Dear Statalist members,
      >
      >      I would like to ask you a question regarding the result of a

Hausman

test.
      >
      >      My question is, with this result, if I conclude that I have

no

problem of


      >      endogeneity; in other words, I have no endogenous variable?
      >
      >      I followed these steps:
      >      1. regress (OLS) to get a residual
      >      2. predict weak_rest1
      >      3. regress (OLS) using weak_rest1
      >      4. regress 2sls using IV
      >
      >      Here is the result of the t test of the residual:
      >      . test weak_res1
      >
      >       ( 1)  weak_res1 = 0
      >
      >             F(  1,   355) =    3.34
      >                  Prob>      F =    0.0686
      >
      >      With is result, can I conclude that no endogeneity problem?
      >
      >      Thank you very much.
      >
      >      Best regards,
      >      Hoang Dinh Quoc
      >
      >
      >
      >
      >      *
      >      *   For searches and help try:
      >      *   http://www.stata.com/help.cgi?search
      >      *   http://www.stata.com/support/statalist/faq
      >      *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*

*   For searches and help try:

*   http://www.stata.com/help.cgi?search

*   http://www.stata.com/support/statalist/faq

*   http://www.ats.ucla.edu/stat/stata/

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of John

Antonakis

Sent: Thursday, April 19, 2012 8:42 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: St: interpret the result of Hausman test

Do:

hausman one two, sigmamore

What does that give? If the hausman test is still NPD try:

ivreg2 y (x = z), endog(x)

Also, did you try it in sem as I suggested?

If the p value of the endogeneity test is<     .05 then x is endogenous.

However, if your sample is small the test might not have much power (so
I would be worried about endogeneity if<     .10). If you have good

reason

to believe that x is endogenous then the iv estimator should be

retained.

HTH,
J.

__________________________________________

Prof. John Antonakis
Faculty of Business and Economics
Department of Organizational Behavior
University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny
Switzerland
Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305
http://www.hec.unil.ch/people/jantonakis

Associate Editor
The Leadership Quarterly
__________________________________________


On 19.04.2012 10:39, Hoang Dinh Quoc wrote:

Dear Prof. Antonakis,

Thank you very much for your quick support.

I followed your suggestion:
"reg y x
est store one
ivregress 2sls y (x=z)
est store two
hausman one two"

And I got this result:

Test:  Ho:  difference in coefficients not systematic

                       chi2(1) = (b-B)'[(V_b-V_B)^(-1)](b-B)
                               =        3.31
                     Prob>chi2 =      0.0687
                     (V_b-V_B is not positive definite)

With is result, can I conclude that no endogeneity problem?

Thanks,
Best,
Hoang Dinh Quoc



-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of John

Antonakis

Sent: Thursday, April 19, 2012 3:23 PM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: St: interpret the result of Hausman test

Hi:

I am not quite sure what you have done here.

If you want to do this "by hand" do an augmented regression:

http://www.stata.com/support/faqs/stat/endogeneity.html

Else, use the -endog- option in the user-written program, ivreg2,
available from ssc (i.e., ssc install ivreg2, replace), e.g. (for
dependent variable y, endogenous regressor x, and instrument z):

ivreg2 y (x = z), endog(x).

Or do the usual hausman test via Stata, e.g.,

reg y x
est store one
ivregress 2sls y (x=z)
est store two
hausman one two

Finally, you can do this in the new Stata command, -sem- using maximum
likelihood:

sem (y<-x) (x<-z), cov(e.y*e.x)

The test of the correlation between the disturbances is the Hausman
test, as we explain in detail here:

Antonakis, J., Bendahan, S., Jacquart, P.,&      Lalive, R. (2010). On
making causal claims: A review and recommendations. The Leadership
Quarterly, 21(6). 1086-1120.
http://www.hec.unil.ch/jantonakis/Causal_Claims.pdf

For more basic explanations see:

Antonakis, J., Bendahan, S., Jacquart, P.,&      Lalive, R. (submitted).
Causality and endogeneity: Problems and solutions. In D.V. Day (Ed.),
The Oxford Handbook of Leadership and Organizations.
http://www.hec.unil.ch/jantonakis/Causality_and_endogeneity_final.pdf


HTH,
J.

__________________________________________

Prof. John Antonakis
Faculty of Business and Economics
Department of Organizational Behavior
University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny
Switzerland
Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305
http://www.hec.unil.ch/people/jantonakis

Associate Editor
The Leadership Quarterly
__________________________________________


On 19.04.2012 10:14, Hoang Dinh Quoc wrote:
      >      Dear Statalist members,
      >
      >      I would like to ask you a question regarding the result of a

Hausman

test.
      >
      >      My question is, with this result, if I conclude that I have

no

problem of

      >      endogeneity; in other words, I have no endogenous variable?
      >
      >      I followed these steps:
      >      1. regress (OLS) to get a residual
      >      2. predict weak_rest1
      >      3. regress (OLS) using weak_rest1
      >      4. regress 2sls using IV
      >
      >      Here is the result of the t test of the residual:
      >      . test weak_res1
      >
      >       ( 1)  weak_res1 = 0
      >
      >             F(  1,   355) =    3.34
      >                  Prob>      F =    0.0686
      >
      >      With is result, can I conclude that no endogeneity problem?
      >
      >      Thank you very much.
      >
      >      Best regards,
      >      Hoang Dinh Quoc
      >
      >
      >
      >
      >      *
      >      *   For searches and help try:
      >      *   http://www.stata.com/help.cgi?search
      >      *   http://www.stata.com/support/statalist/faq
      >      *   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index