Date:         Sun, 20 Sep 2009 10:44:11 -0700
Reply-To:     Bobs <fabianezte@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Bobs <fabianezte@GMAIL.COM>
Organization: http://groups.google.com
Subject:      matrix hessian
Comments: To: sas-l@uga.edu
Content-Type: text/plain; charset=ISO-8859-1

Hello all,

I'm trying to set a non-linear mixed model using the NLMIXED but in
the log window is leaving the following message below:

WARNING: The final Hessian matrix is not positive definite, and
therefore the estimated covariance matrix is not full rank and may be
unreliable.  The variance of some parameter estimates is zero or some
parameters are linearly related to other parameters.

What should I do?
I do not know what is wrong in the commands

What I want is to model the variability between and within animals
that are longitudinal data where several observations were made in the
same animal.

DATA A1;
INPUT cows treat$  weight type$  age;
DATALINES;
 1 TA  333 P12   365
 1 TA  453 P18   603
 1 TA  462 P24   703
 1 TA  256 PD    253
 1 TA   31 PN      0
 1 TA  445 PVD   958
 1 TA  380 PVE   407
 1 TA  462 PVP   703
 1 TA  432 PVS   540
 1 TS  300 P12   355
 1 TS  441 P18   593
 1 TS  235 PD    226
 1 TS   41 PN      0
 1 TS  431 PVD  1018
 1 TS  386 PVE   495
 1 TS  486 PVP   815
 1 TS  417 PVS   530
 3 NI  299 P12   393
 3 NI  362 P18   589
 3 NI  416 P24   742
 3 NI  216 PD    238
 3 NI   33 PN      0
 3 NI  457 PVD  1323
 3 NI  416 PVE   758
 3 NI  462 PVP  1069
 3 NI  449 PVS   855
;
proc print data=A1;run;

proc sort data=A1;
BY treat COWS;
RUN;

proc nlmixed data=A1 corr cov;
*by gg cows;
Parms
A=470
b=0.95
k=.0016
s2e=100 s2u1=1000 s2u2=900 c=0.3;
bounds b>0, k>0, 300<=A<=1000;
brody = (A+e1)*(1-B*EXP(-(K+e2)*AGE));
model weight ~ normal (brody,s2e);
random e1 e2 ~ normal([0,0],[se13,c,se23]) subject=COWS;
run;

Please help me!