Here is an example of some output from -probitiv- using garbage
data:


* first we will run a regular probit with the left hand side variable
* being "upst" and the right hand side variables being "fsmax",
* "fsnet", and "maxpay":
. probit upst fsmax fsnet maxpay

Iteration 0:   log likelihood = -9090.8142
Iteration 1:   log likelihood =  -8077.609
Iteration 2:   log likelihood = -8052.7507
Iteration 3:   log likelihood = -8052.3896
Iteration 4:   log likelihood = -8052.3803

Probit estimates                                  Number of obs   =      13209
                                                  LR chi2(3)      =    2076.87
                                                  Prob > chi2     =     0.0000
Log likelihood = -8052.3803                       Pseudo R2       =     0.1142

------------------------------------------------------------------------------
    upst |      Coef.   Std. Err.       z     P>|z|       [95% Conf. Interval]
---------+--------------------------------------------------------------------
   fsmax |  -.0006839   .0000379    -18.068   0.000      -.0007581   -.0006097
   fsnet |   .0001877   .0000135     13.861   0.000       .0001612    .0002142
  maxpay |   .0002746   6.99e-06     39.270   0.000       .0002609    .0002883
   _cons |  -.7670535   .0288786    -26.561   0.000      -.8236546   -.7104524
------------------------------------------------------------------------------

* next we will run -probitiv- with the left hand side variable still
* being "upst", the exogenous right hand side variables being
* "fsnet" and "maxpay". The endogenous right hand side variable is
* "fsmax", and the instrumental variable is "fsgross":
. probitiv upst, exog(fsnet maxpay) endog( fsmax ) iv(fsgross)
NOTE: ESTIMATES NOT CONSISTENT WHEN ENDOG VARIABLE IS BINARY!!
CODE STILL RUNS, WHICH IS A PROBLEM.......
(option xb assumed; fitted values)
(7140 missing values generated)
(0 real changes made)
(7140 missing values generated)
(0 real changes made)


--------
--------
TEMPVAR: __000000 fsmax ;  probitiv used    6069 observations

------------------------------------------------------------------------------
    upst |      Coef.   Std. Err.       t     P>|t|       [95% Conf. Interval]
---------+--------------------------------------------------------------------
__000000 |  -.0015201   .0006926     -2.195   0.028      -.0028779   -.0001623
   fsnet |   .0005542   .0002709      2.046   0.041       .0000232    .0010852
  maxpay |    .000207   .0000102     20.360   0.000       .0001871     .000227
__000001 |  -.0005575   .0000695     -8.025   0.000      -.0006937   -.0004213
   _cons |   -1.67872   .6695667     -2.507   0.012      -2.991308   -.3661312
------------------------------------------------------------------------------
ln_L     |  -3659.929916491127 . . . . .


* Things to note about -probitiv-'s output: 
* 
* 1) The two lines (all in caps) about binary endogenous variables
* will always display -- they are there as a reminder.
*
* 2) The lines with stuff in parens are stata's output from the
* "first-stage" and prediction parts of the estimates. They are just
* things I forgot to put a -quietly- around. Ignore them unless you
* are hacking the code and trying to debug.
*
* 3) The estimated IV coefficient on the endogenous right hand side
* variable is given in this case by the row in the display matrix
* beginning with "__000000". The reason is that to run the second
* stage, -probitiv- first runs a linear regression of the form:
*
* . reg fsmax fsnet maxpay fsgross
*
* after which the predicted values of both "fsmax" and the residual
* from this regression are created. Those residuals are then used in 
* probit estimation of the form:
*
* . probit upst fsmaxhat fsnet maxpay ehat
*
* where "fsmaxhat" and "ehat" are the predicted values just
* described. 
*
* 4) Hence in the example above, the variable "__000000" is the
* predicted value of "fsmax" used in the second stage probit. This is
* why the line above the coefficients table says 
*
* "TEMPVAR: __000000 fsmax"
*
* This output is telling you the name of the variable that represents
* predicted values of the variable "fsmax".
*
* 5) In the example above, the variable "__000001" is the predicted
* value of the first-stage residual. It is an ancillary parameter and
* can be ignored. See Newey's article if you are still wondering.
*
* 6) Don't bother trying to find out stuff about the predicted-values
* variables (in this example "__000001" and "__000000"). These are
* internal, temporary variables created by -probitiv-, and they are
* destroyed as soon as -probitiv- finishes.
*
* 7) While -probitiv- spits out a log likelihood statistic from the
* second-stage probit, I am not sure it is meaningful (since there are
* predicted values on the right hand side). Maybe Newey's article says
* something about this.


Please note that everything above applies to -tobitiv-. Note also that
the second-stage tobit assumes a lower limit (invariant across
observations). If you need to change this, just go to line #69 of
tobitiv.ado and make the necessary change.