st: nl using estimated coefficients

[Kumiko Imai <ki25@cornell.edu>]

Hello,

Does anyone have any ideas how to fix the following problem?  After 
estimating a logit I use the estimated coefficients in an nl routine but I 
get the "type mismatch" error.  I suspected the problem lies in the use of 
scalars, so I first svmat the coefficients and then use those instead of 
the scalars, but then I get the "RHS variables missing" error.

Thanks very much,

Kumiko

------------------------------------
. logit anycost age

Iteration 0:   log likelihood = -2843.9766
Iteration 1:   log likelihood =  -2843.968
Iteration 2:   log likelihood =  -2843.968

Logit estimates                        Number of obs   = 5055
LR chi2(1)      =       0.02
Prob > chi2     =     0.8954
Log likelihood =  -2843.968           Pseudo R2 = 0.0000


anycost   Coef.   Std. Err.    z    P>z   [95% Conf. Interval]

age   -.0004165   .0031672    -0.13   0.895    -.0066242   .0057911
_cons  1.112459   .119838     9.28   0.000      .877581    1.347337


. scalar b10=_b[_cons]

. scalar b11=_b[age]

. di b10 b11
1.1124592-.00041651

.
. program define nlfcn
1.         if "`1'"="?"{
2.         global S_1 "b20 b21"
3.         global b20=.1
4.         global b21=.1
5.         exit
6.         }
7.         replace `1'=exp((b10+$b20)+(b11+$b21)*age)/(1+exp(b10+b11*age))
8. end

.
. nl fcn cost
type mismatch
nlfcn refused query, rc=109
r(109);

end of do-file
r(109);

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/