data stat.ces;
input l k logq @@;
cards;
.228 .802 -1.359 .258 .249 -1.695
.821 .771 .193 .767 .511 -.649
.495 .758 -.165 .487 .425 -.270
.678 .452 -.473 .748 .817 .031
.727 .845 -.563 .695 .958 -.125
.458 .084 -2.218 .981 .021 -3.633
.002 .295 -5.586 .429 .277 -.773
.231 .546 -1.315 .664 .129 -1.678
.631 .017 -3.879 .059 .906 -2.301
.811 .223 -1.377 .758 .145 -2.270
.050 .161 -2.539 .823 .006 -5.150
.483 .836 -.324 .682 .521 -.253
.116 .930 -1.530 .440 .495 -.614
.456 .185 -1.151 .342 .092 -2.089
.358 .485 -.951 .162 .934 -1.275
;
run;
proc nlin data=stat.ces;
parms b0=1 a=-1 d=.5 r=-1;
lr=l**r;
kr=k**r;
z=d*lr+(1-d)*kr;
model logq=b0+a*log(z);
der.b0=1;
der.a=long(z);
der.d=(a/z)*(lr-kr);
der.r=(a/z)*(d*log(l)*lr+(1-d)*log(k)*kr);
run;
报错为：
ERROR: Cannot find a library containing subroutine LONG.
ERROR: no dependent variable specified.作者: shiyiming 时间: 2008-10-26 10:39 标题: Re: 求高人指点NLIN的问题，程序报错，不知道为什么 it's hard to find one thing i know for certain!
it's just a typo. the partial derivative of logq with respect to a is 'log(z)' instead of 'long(z)'.
sorry i may appear too excited, but this world is full of uncertainties! a typo is one of the few only exceptions.作者: shiyiming 时间: 2008-10-26 10:46 标题: Re: 求高人指点NLIN的问题，程序报错，不知道为什么 Think you very much, you are right .