[Limdep Nlogit List] Using maximize
William Greene
wgreene at stern.nyu.edu
Wed Nov 7 06:49:15 EST 2007
Bob. You are taking the log of Phi(...). The argument being
sent to the PHI function is (apparently) less than (roughly)
-10, and it is returning zero. Or, the argument is very large,
and it is returning a 1.0, which then transformed to 1-Phi()
where logs appear once again. Finally, your parameter
GE is not constrained to be positive, and you are taking logs of it.
/Bill
----- Original Message -----
From: "Robert Nakosteen" <nakosteen at som.umass.edu>
To: limdep at limdep.itls.usyd.edu.au
Sent: Tuesday, November 6, 2007 2:32:15 PM (GMT-0500) America/New_York
Subject: [Limdep Nlogit List] Using maximize
Greetings,
I have written the following program in Limdep, and created a small data set
with which to test it.
? Recursive model with first stage OLS and second stage Probit: Hausman and
Wise 1979
read; nrec=31; nvar=10; names=distance, hearn, heduc, hage, unem, child,
wage, weduc, wearn, Dummy; maxvar=50$
0 100 10 30 8 5 18 9 80 0
0 118 11 36 6 3 19 9 75 0
0 120 11 40 12 2 19 10 90 1
0 125 12 19 11 6 20 10 95 0
0 90 13 26 7 8 20 12 98 0
0 110 12 50 5 4 22 11 90 1
0 116 14 65 3 2 22 12 90 1
0 119 13 21 11 2 25 12 100 1
0 140 10 40 9 1 25 13 89 1
0 125 9 28 8 6 27 13 110 0
0 129 13 70 8 7 30 13 120 0
0 127 12 29 6 5 32 13 118 1
0 129 16 40 11 6 32 14 116 0
0 142 12 45 6 3 35 14 119 0
0 130 11 38 4 2 39 15 122 1
5 140 10 40 9 5 42 14 131 1
5 141 12 29 13 3 43 14 129 1
5 120 13 30 14 3 43 15 127 1
5 136 14 22 9 2 45 16 126 0
5 150 14 28 8 1 49 17 130 1
10 160 14 36 11 1 52 17 140 1
10 168 16 29 14 3 55 18 135 1
10 170 16 40 13 1 55 18 133 1
10 160 15 33 16 4 58 17 136 1
10 159 13 37 9 2 59 16 139 0
15 170 15 28 14 0 60 15 150 1
15 170 15 28 14 0 62 16 149 1
15 200 17 40 16 1 65 17 148 0
15 180 16 22 14 3 67 14 148 0
15 178 15 20 13 2 67 13 146 1
15 190 19 27 10 4 68 12 130 1
namelist; x = one,hage,heduc; ? variables on the rhs of the earnings
equation
z = one,wearn,child,unem$ ? exog variables on the right hand side
of the Probit equation
dstats; rhs=*$
? single equation estimates to get starting values - use a guess for the
covariance
regress; lhs=hearn; rhs=x$
Probit; lhs = dummy; rhs=z,hearn$
?
maximize; fcn = bx = b0 + b1*hage + b2*heduc
| gz = g0 + g1*wearn + g2*child + g3*unem
| P1 = a*b0 + a*b1*hage + a*b2*heduc + gz
| P2 = a*se*se
| P2 = P2 + sew
| P2 = P2/(se*se)
| P2a = P2*(hearn-bx)
| P31 = a*a*se*se
| P31 = P31 + 2*a*sew +1
| P32 = (P2 + sew)^2
| P32 = P32/(se*se)
| P3 = sqr(P31-P32)
| PhiA = Phi((P1 + P2a)/P3)
| ge = N01((hearn - bx)/se)/se
| L1 = log(PhiA)
| L2 = log(1-PhiA)
| L3 = log(ge)
| dummy*(L1 + L3) + (1-dummy)*(L2 + L3);
labels = b0,b1,b2,g0,g1,g2,g3,a,se,sew;
start = 43.6,-0.25,8.07,2.00,0.01,-0.40,0.03,-0.02,20,0.10$
Stop$
I get the following error messages when I run the program:
Error: 590: Obs.= 1 Cannot compute function: Logminus
Error: 137: Iterations: function not computable at crnt. trial estimate
Error: 802: Cannot compute function at start values. Exit status=4.
Function= .19226648919D+02, at entry, .00000000000D+00 at exit
Error: 590: Obs.= 1 Cannot compute function: Logminus
Error: 137: Iterations: function not computable at crnt. trial estimate
Error: 143: Models - estimated variance matrix of estimates is singular
Error: 447: Current estimated covariance matrix for slopes is singular.
Since nothing I am logging takes on a negative value, I am stymied.
Can anyone help?
Robert A. Nakosteen
Professor
Isenberg School of Management
University of Massachusetts/Amherst
Amherst, MA 01003
Telephone: 413-545-5687
FAX: 413-545-3858
Fin-opmgtfaculty mailing list
Fin-opmgtfaculty at som.umass.edu
https://list.umass.edu/mailman/listinfo/fin-opmgtfaculty
_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au
--
Professor William Greene
Department of Economics
Stern School of Business
New York University
44 West 4th St., Rm. 7-78
New York, NY 10012
http://www.stern.nyu.edu/~wgreene
212.998.0876
More information about the Limdep
mailing list