Dear R-users
My problem is the following:
1) Suppose we have the data test.txt:
Data has form: Subject age test (=binary variable)
1 0.6 0
2 2 1
3 0.9 1
...
2) I would like solving the following differential equations:
dA/dt = u*A + v*B
dB/dt = -v*B -u*A
in addition, the parameter
u=k*t
where k is a constant and t is the time.
How could I implement that into nlmeODE?
I tried the following procedure:
a) Apply the nlmeODE function
dataModel=groupedData(formula=test~age|Subject,
data=read.table("E:/test.txt",header=T),
labels=list(x="Age",y="testpositiv"),
units=list(x="(year)",y="(+/-)"))
modelform <- list(DiffEq=list(
dSdt = ~ u*I-b*S,
dIdt = ~ -u*I+b*S,
dbdt = ~ -k ),
ObsEq=list(
c1 = ~ S,
c2 = ~ 0,
c3 = ~ 0),
States=c("S","I","b"),
Parms=c("u","k","start"),
Init=list("start",0,0.3))
Modeleasy <- nlmeODE(modelform,dataModel)
b) as far as good, now the problem if I run an estimation
Modeleasy.nlme <- nlme(test ~ Modeleasy(u,k,start,age,Subject),
data = dataModel, fixed=u+k+start~1, random = pdDiag(start+u~1),
start=c(u=5.7,k=-2,start=-5.9),
control=list(msVerbose=TRUE,tolerance=1e-1,pnlsTol=1e-1,msTol=1e-1),
verbose=TRUE)
plot(augPred(Modeleasy.nlme,level=0:1))
=> I get always an error message "leading Minor of degree 2 is not
positiv definit"
Hope anybody can help me or propose an alternative method to solve
that problem.
Thanks a lot. Best regs...
--
Dominik Heinzmann
Master of Science in Mathematics, EPFL
Ph.D. student in Biostatistics
Institute of Mathematics
University of Zurich