Date:         Thu, 23 Sep 2010 09:56:02 -0400
Reply-To:     FRP3@PITT.EDU
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         SUBSCRIBE SAS-L Ameribrit <FRP3@PITT.EDU>
Subject:      NLMIXED "Observation error"
Content-type: text/plain; charset=iso-8859-1

HI Guys: I was hoping someone could help me with an issue in PROC NLMIXED.
I receive emails from listserv and tried to send this directly to dale but
I am not sure if thats the protocol for submitting issues.

Anyway I am currently trying to model survival with hazard function:

h(t)=alpha(0)*exp(XB+r1*u1+r2*u2+r3*(u1+u2*t)+u3.

This is from a paper by carlin et al where this is the survival piece of a
joint model. I am assumng an exponential model for now so alpha0 is some
constant say bs0.

To complete the likelihood  L= h(t)*S(t) I need the survival function.
This is what I get after integrating

S(t) =exp( - A /B *exp(Bt-1)

A= alpha(0)*exp(XB+(r1+r3)*u1+r2*u2+u3)
B= r3*u2

So the code I am trying in SAS is as follows:


proc nlmixed data=full1 absgconv=1E-05 qpoints=3 ;

parms /data=seperateestimates;

/*survival part*/
if(last=1) then do;

linpred=bs0+bs1*X+r1*u1+r2*u2+r3*(u1+u2*survtime)+u3;        /*linear
predictor*/

A=exp(bs0+bs1*X+(r1+r3)*u1+r2*u2+u3);
B=r3*u2;
C=A/B;

h_t= exp(linpred); /*hazard*/
  S_t= exp(-C*(exp(B*survtime)-1)); *exponential survival function;
      f_t= h_t*S_t; *Likelihood = h(t)*S(t);

     llsurv = (censored=0)*log(f_t) + (censored=1)*log(S_t);
*contribution to the likelihood;

end; else llsurv=0;

model last ~ general(llsurv);

random u1 u2 u3   ~
normal([0,0,0],[exp(2*s11),0,exp(2*s22),0,0,exp(2*s33)]) subject=id;

estimate "s11" exp(2*s11);
estimate "s33" exp(2*s33);
estimate "se"  exp(se);
run;

But the error I keep getting is:
 Execution error for observation 3.
NOTE: PROCEDURE NLMIXED used (Total process time):
      real time           0.40 seconds
      cpu time            0.32 second


Any help from you guys would be appreciated. I know I may have left out a
lot of info so please let me know if you need further clarification.
thanks
Francis