Re: st: Estimating SEM with Dummy

[John Antonakis <John.Antonakis@unil.ch>]

Hi:

One more thing. All exogenous variables should go in all upstream 
equations. Thus, you should better do:

sem (mmr ms lending era1-era4 cpi cps -> cps) (ms era1-era4 cpi cps-> 
cpi) (era1-era4 cpi cps -> gdp_growth), cov(e.cpi*e.cps 
e.cpi*e.gdp_growth e.cps*e.gdp_growth)

See: Baltagi, B. H. (2002). Econometrics. New York: Springer.

Best,
J.

__________________________________________

John Antonakis
Professor of Organizational Behavior
Director, Ph.D. Program in Management

Faculty of Business and Economics
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
Organizational Research Methods
__________________________________________

On 30.03.2014 23:35, John Antonakis wrote:
> Are era1-era5 your dummy variables already? Seems to me that they may
be. If so, you may have fallen in the "dummy variable trap." You need 
k-1 dummies. You have 5 categories, thus need 4 dummies. You should be 
estimating (note you can put more than one regressor in an equation):
>
> sem (mmr ms lending -> cps) (ms -> cpi) (era1-era4 cpi cps -> gdp_growth)
>
> Now, because cps and cpi are endogenous regressors, you cannot assume
that the cross-equation disturbances are orthogonal. To obtain 
instrumental variable estimates you should do this:
>
> sem (mmr ms lending -> cps) (ms -> cpi) (era1-era4 cpi cps ->
gdp_growth), cov(e.cpi*e.cps e.cpi*e.gdp_growth e.cps*e.gdp_growth)
>
> The Hausman test of endogeneity will be a 2 DF test of the latter two
covariances = 0, that is:
>
> test (_b[cov(e.cpi*e.gdp_growth):_cons] = 0)
(_b[cov(e.cps*e.gdp_growth):_cons] = 0)
>
> To better understand why you must do this see:
>
> http://www.youtube.com/watch?v=dLuTjoYmfXs
>
> For background reading see:
>
> 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
>
> [If you wish, refer to the following “prequel” paper, which is really
a more basic introduction to endogeneity]:
>
> Antonakis, J., Bendahan, S., Jacquart, P., & Lalive, R. (2014).
Causality and endogeneity: Problems and solutions. In D.V. Day (Ed.), 
The Oxford Handbook of Leadership and Organizations (pp. 93-117). New 
York: Oxford University Press.
> http://www.hec.unil.ch/jantonakis/Causality_and_endogeneity_final.pdf
>
> HTH,
> John.
>
>
>
> __________________________________________
>
> John Antonakis
> Professor of Organizational Behavior
> Director, Ph.D. Program in Management
>
> Faculty of Business and Economics
> 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
> Organizational Research Methods
> __________________________________________
>
> On 30.03.2014 23:20, Theophilus Dapel wrote:
> > Dear Billy,
> > Thank you so much for the clarification.
> >
> > I have been building a SEM for path analysis (system of equations).
The dependent variables are x1, x2 and x3 (continuous variables).The 
depending variable is y (continuous variable). I wish to find the impact 
of the independent variables through intermediate variables z1 and z2 
(also continuous variables). And five dummies: era1, … and era5.
> >
> > Here is the model
> > xi: sem (mmr -> cps, ) (ms -> cps, ) (ms -> cpi, ) (lending -> cps,
) (cps -> gdp_growth, ) (cpi -> gdp_growth, ) (i.era1 -> gdp_growth, ) 
(i.era2 -> gdp_growth, ) (i.era3 -> gdp_growth, ) (i.era4 -> gdp_growth, 
) (i.era5 -> gdp_growth, ), standardised
> >
> > On 30 Mar 2014, at 22:03, William Buchanan
<william@williambuchanan.net> wrote:
> >
> >> John's example was a simple linear regression.  Again, you really
should provide more information about exactly what you are trying to do. 
 The error message you received means that the values Stata chose by 
default for the starting values weren't good. You can either try 
specifying starting values for the model manually or it's possible that 
the -difficult- option for MLE algorithm could alleviate the problem. 
However, if you have a series of dummy variables (e.g., already coded as 
0|1), then the xi: prefix isn't doing anything.  If, instead, you have a 
series of nominal variables with several categories you could use the 
generate option of -tabulate- to create your dummy variables.
> >>
> >> HTH,
> >> Billy
> >>
> >> Sent from my iPhone
> >>
> >>> On Mar 30, 2014, at 15:42, Theophilus Dapel <dapelzg82@gmail.com>
wrote:
> >>>
> >>> Dear John,
> >>> I applied the option you gave, thanks.
> >>>
> >>> But I got an error message: initial values not feasible
> >>>> On 30 Mar 2014, at 17:07, John Antonakis
<John.Antonakis@unil.ch> wrote:
> >>>>
> >>>> Hi:
> >>>>
> >>>> In terms of controls, that is, right hand side variables. Just
do as you would normally:
> >>>>
> >>>> xi: sem (y <- x z i.dummy)
> >>>>
> >>>> Stata will generate k-1 dummies for you, like with other estimators.
> >>>>
> >>>> Best,
> >>>> J.
> >>>>
> >>>> __________________________________________
> >>>>
> >>>> John Antonakis
> >>>> Professor of Organizational Behavior
> >>>> Director, Ph.D. Program in Management
> >>>>
> >>>> Faculty of Business and Economics
> >>>> 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
> >>>> Organizational Research Methods
> >>>> __________________________________________
> >>>>
> >>>>> On 30.03.2014 16:38, Theophilus Dapel wrote:
> >>>>> Dear User of Stata,
> >>>>> Any assistance on how to estimate a SEM with dummies in the
list of control variables?
> >>>>>
> >>>>> Thanks,
> >>>>> Dapel
> >>>>> *
> >>>>> *   For searches and help try:
> >>>>> *   http://www.stata.com/help.cgi?search
> >>>>> * http://www.stata.com/support/faqs/resources/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/faqs/resources/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/faqs/resources/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/faqs/resources/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/faqs/resources/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/faqs/resources/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/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/