是不是和“ observation 1.”这个变量有关？

It could be a illegal math operations . For example y=srqt(-1) or z=log(-8) , see an example below.


data t1;
   r=0.1;
   do i=1 to 100;
      x=rannor(123); x2=(x+r*rannor(123))/sqrt(1+r**2);
          y=0.001+1*x+1*x2>rannor(123);
          output;
        end;
run;

proc print;run;

proc nlmixed data=t1 hess cov;
parms a=0 b=0 c=0;

z=log(x);

xbeta=a+b*x+c*x2;
if y=1 then p=probnorm(xbeta);
else p=1-probnorm(xbeta);
loglik=log(p);
model y  ~ general (loglik);
run;