From: Nilesh Timilsina on
Dear Sas Experts

I am trying to run a individual tree diameter growth model using Proc
Nlmixed. Below is the code i am trying to run. But I always get the message,
"Quadrature accuracy of 0.000100 could not be achieved with 31 points. The
achieved accuracy was 0.00044", if i run the model with method=gauss. If the
same model is used with method=firo i get the message, "Floating point
overflow, termination due to floating point exception. I am new to this non
linear mixed model thing. I would really appreciate the help to solve this
problem

proc
nlmixed data = test method=firo;
parms
a0=0.1141 a2=-0.00100 a3=-0.01453 /*a4=-0.00042 c0=0.1094*/ c1= 0.002151
log_sd=0 log_s2u1=1 log_s2u2=1
z=0;
s2 = exp(2*log_sd);
s2u1=exp(2*log_s2u1);
s2u2=exp(2*log_s2u2);
rho = (exp(2*z)-1)/(exp(2*z)+1);
cv1 = rho*exp(log_s2u1+log_s2u2);
b= a0 + (a2+u1)*totht + (a3+u2)*sum_SI;
c= c1*spi2;
pred = (b/(b-c))*(exp(-c*curdbh)-exp(-b*curdbh));
MODEL
ldbhgr3 ~ normal(pred, S2); RANDOM
u1 u2 ~ normal([0,0],[s2u1,cv1,s2u2]) SUBJECT=fk_sdi OUT=out_Rand;
PREDICT pred out=preds;
run;

Thanking you
Nilesh Timilsina