[WinBUGS] Fixed multivariate meta-analysis problems

Gwen

New Member
#1
Hello !

I am currently a trainee for a health economic consultancy firm. I am comparing different outcomes for a disease. For this purpose, I have done a fixed multivariate meta-analysis on WinBugs, but there is a problem. Actually, The model is syntactically correct, the data were loaded but in the moment, I am updating the model, there is an error "undefined real result", I do not undestand why. Could you help me ?

Thank you a lot.

Here there are the data, the program and also the initial values.

Model{
for (i in 1: Num){
ln_OR[i,1:2] ~ dmnorm(d[1:2],weight[(2*i-1):(2*i),1:2])
}
for (j in 1: 2){
d[j] ~ dnorm(0,1)
OR[j] <-exp(d[j])
VE[j] <- (1-OR[j])*100
}

}


list(Num=5.00000E+00, ln_OR= structure(.Data= c(-1.46712E+00, -1.52210E+00, -1.05361E-01, -5.82799E-01, -7.94930E-01, -1.88137E+00, 8.63773E-01, -5.67984E-01, -6.76037E-01, -1.44101E+00), .Dim=c(5, 2)), weight= structure(.Data= c(-9.48602E+00, 3.28027E+00, 3.28027E+00, -3.01109E-01, 2.34239E+00, -3.20671E-01, -3.20671E-01, 3.73759E-01, 2.61258E+00, -7.61214E-01, -7.61214E-01, 9.23839E-01, 6.72839E-01, -9.18010E-02, -9.18010E-02, 3.41275E-01, -1.42907E+00, 2.71617E+00, 2.71617E+00, -5.82852E-01), .Dim=c(10, 2)))

list(d=c(-0.5,-0.5))
list(d=c(0,0))
list(d=c(0.5,0.5))