switch_probit

[Pat OBrien]

Hello Statalist,

I am trying to use switch_probit for an endogenous switching model. I once used it and have a regression saved in a do file from a couple years ago, but now Stata won't even recognize the command, and I can't find any help online on how to install it again. Appreciate any help.

[Andrew Musau]

. net sj 11-3 st0233

----------------------------------------------------------------------------------------------------------------------------------------
package st0233 from http://www.stata-journal.com/software/sj11-3
----------------------------------------------------------------------------------------------------------------------------------------

TITLE
SJ11-3 st0233. Maximum likelihood estimation of...

DESCRIPTION/AUTHOR(S)
Maximum likelihood estimation of endogenous switching
probit model
by Michael Lokshin, The World Bank
Zurab Sajaia, The World Bank
Support: [email protected]
After installation, type help switch_probit

INSTALLATION FILES (type net install st0233)
st0233/switch_probit.ado
st0233/switch_probit.sthlp
st0233/switch_probit_d2.ado
st0233/switch_probit_p.ado

ANCILLARY FILES (type net get st0233)
st0233/switch_probit_example.do
st0233/switch_probit_example.dta
----------------------------------------------------------------------------------------------------------------------------------------

So

net install st0233

[Pat OBrien]

Andrew --

Thank you. Although when I use that I get:

file http://fmwww.bc.edu/repec/bocode/s/st0233.pkg not found
could not load st0233.pkg from http://fmwww.bc.edu/repec/bocode/s/

It would seem that this package isn't available anymore at that university's site. But looking at http://fmwww.bc.edu/repec/bocode/s/ I can see a couple similarly named packages like switchoprobit. But I don't see much help on figuring it out.

Is there a Stata-authored command that should do the same thing? That is, one that involves a binary endogenous treatment equation and main, also binary equation?

Pat

[Andrew Musau]

This command is available from the Stata Journal and not SSC. Did you type

net install st0233

?

[Pat OBrien]

Yes I did. So why would this lead me to a third party site?

. net install st0233
file http://www.stata.com/st0233.pkg not found
could not load st0233.pkg from http://www.stata.com/
r(601);

[Pat OBrien]

Sorry, that was confusing. Earlier when entered the code just as you put it it came back with what I posted with the bc.edu URLs. I restarted and tried again, and this time it goes to stata.com as I just posted prior to this, but as you can see it still can't load.

[Andrew Musau]

I have no problems installing the program.

.  net install st0233
checking st0233 consistency and verifying not already installed...
installing into C:...
installation complete.

. which switch_probit
C:\...
*! version 1.0 14Jan2009 M. Lokshin, Z. Sajaia

Do you have a Windows PC? Try

findit switch_probit

and see if you can install clicking on the SJ links provided.

[Pat OBrien]

Yes, that worked. Thank you!

[Kumari Gunjan]

Hello Sir,
I have very basic question in understanding the result by switch_probit. I have given the command as follows:

 switch_probit FARM $xlist, select(program $selection)

and result is as given below.

Switching probit model    Number of obs    =     21,924
    Wald chi2(15)    =    5669.24
Log likelihood = -21350.332    Prob > chi2    =     0.0000    
Coef.   Std. Err.    z    P>z    [95% Conf. Interval]
        
program     
head_edu    .0371164    .002472    15.01   0.000    .0322714    .0419614
head_age    .0098915   .0009282    10.66   0.000    .0080723    .0117107
land_owned    .0010394   .0001837    5.66   0.000    .0006795    .0013994
_cons   -3.470054   .1939979    -17.89   0.000    -3.850283   -3.089825
        
FARM_1            
age    .1079043   .0086594    12.46   0.000    .0909323    .1248764
age2   -.0012392   .0001117    -11.09   0.000    -.0014581   -.0010202
edu_years   -.0326016   .0035525    -9.18   0.000    -.0395643   -.0256389
_cons   -3.387048   .3727443    -9.09   0.000    -4.117613   -2.656482
        
FARM_0            
age    .0918237    .010393    8.84   0.000    .0714537    .1121936
age2   -.0010055   .0001352    -7.44   0.000    -.0012705   -.0007405
edu_years   -.0188371   .0041628    -4.53   0.000    -.0269961   -.0106781
_cons   -4.336778   .4242766    -10.22   0.000    -5.168345   -3.505211
        
/athrho1   -.4844595   .0613388        -.6046814   -.3642375
/athrho0   -.2446718   .0682499        -.3784391   -.1109045
        
rho1   -.4498079   .0489283        -.5403724   -.3489411
rho0   -.2399035   .0643218        -.3613511    -.110452      
LR test of indep. eqns. (rho1=rho0=0):chi2(2)    =    64.01  Prob    > chi2 = 0.0000        

My question is that-
1) which test statistics is used to see the significance of rho1 and rho0 .
2) If I want to estimate the treatment effect for various subgroups. Do I need to run the model again and again for various subgroups. Or I can run the model on whole population and predict the treatment effect based on that model for whole individuals. From that predicted probabilities, we can tabulate predicted treatment effect for various groups. Please clarify these doubts . I am struggling with these minor problems from long time.