NOTE: Execution error for an observation in PROC NLMIXED

[SR79]

Hi, 

I am trying to use proc nlmixed on a set of simulated data to check for different models and their AIC/BIC criteria. This attempt below is to model the first one which is: response = y0*exp(-kd*t). 

Below, I have the simulated data, dropped the parameter kd and then try to fit the nlmixed:

data simulated;
call streaminit(9);
do subject =1 to 100;
y0=rand('normal',1,5);
k1=rand('normal', 2,2);
k2=rand('normal',7,2);
kd=rand('normal',2,3);
do t = 1 to 10;
response=y0*exp(-kd*t);
output;
end;
end;
run;

data simulated2;
set simulated;
drop kd;
run;

proc nlmixed data=simulated2 method=gauss;
parms kd=1 delta=0.5 s2=2;
pred=y0*exp(-kd*t);
model response~normal(pred, pred**(2*delta)*s2);
random kd ~ normal(1,2)
subject=subject;
run;

I am getting the following error:

NOTE: Execution error for observation 11.

There's nothing specific about observation 11 which is the first record for subject 2: 

Any thoughts/help on this would be much appreciated!!

Thank you!

[PaigeMiller]

Work through the math. Observation 11 produces an invalid value in the second argument here:

model response~normal(pred, pred**(2*delta)*s2);

[SR79]

Thanks for responding. I updated the code to make sure it's a positive value for variance but now I have the following coming up:

ERROR: Quadrature accuracy of 0.000100 could not be achieved with 31 points. The achieved
accuracy was 1.000000.

Does this mean that the problem doesn't converge?

Thank you. 

[PaigeMiller]

I have not had that happen to me, but "does not converge" is what I think it means.

[SR79]

There was a problem with the input statement, I think I have figured it out. 🙂 Thanks!!

[PaigeMiller]

Please explain what the solution was that you found.

[SteveDenham]

I am in the same boat as @PaigeMiller .  You say the problem was with the input statement.  I don't see such a thing in the code you posted, and as far as I know, PROC NLMIXED doesn't support an "input statement".  So, how did you solve this issue?

SteveDenham