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

st: Re: Simulating Instrumental Variable Probit Model - Addendum

From   Sachin Chintawar <>
Subject   st: Re: Simulating Instrumental Variable Probit Model - Addendum
Date   Tue, 21 Apr 2009 11:08:04 -0600

Dear Statalist Users
Well adding to my earlier program I did find a few silly mistakes that
I have cleaned up. Find a newer version of the program with more
global numobs 500             // sample size N
global numsims "1000"         // number of simulations
set seed 123456789

capture program drop endoprob

program endoprob, rclass
	version 10.1
	drop _all
set obs $numobs
    generate u = rnormal(0)
    generate mu = rnormal(0)
    generate x1 = rnormal(0)
    generate z1 = rnormal(0) 			  // 4 Instruements
    generate z2 = rnormal(0)
    generate z3 = rnormal(0)
    generate z4 = rnormal(0)
    						// Also be written *drawnorm u mu x1 z1 z2 z3 z4
    generate a = 0.5*u
    generate x2 = z1 + z2 + z3 + z4 + a
    regress x2  z1 z2 z3 z4 mu                // endogenous regressor
with four instruments
    predict px2
    generate y = 0.5 + 2*x1 + px2 + mu + u>4    //Reduced Form Equation
    ivprobit y x1 px2
    return scalar b2 = _b[x1]
    return scalar b3 = _b[px2]
    return scalar se2 = _se[x1]								
    return scalar se3 = _b[px2]
simulate b2=r(b2) b3=r(b3) se2r=r(se2) se3r=r(se3), ///
    reps($numsims): endoprob
mean b2 b3 se2r se3r
1. Adding to my earlier problem if we run this program an error occurs
suggesting that there is no endogenous variable  and that I use
-probit- . But this totally negates the question I want to answer -
the Amount of bias as we increase the number of instruments?
Any suggestions would be appreciated

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index