Z is the log-transformed bivariate response x[i,1] is a sequence indicator x[i,2] is a period indicator x[i,3] is a treatment indicator s[g[i]] corresponds to the random subject effect g[i] is a variable indicator of the subject effect Model { for (i in 1:N) { z[i,1:2]~dmnorm(mu[i,1:2],tau1[1:2,1:2]) mu[i,1]<-beta1[1]*x1[i,1]+beta1[2]*x1[i,2]+alpha[1]*x1[i,3]+s[g[i],1] mu[i,2]<-beta2[1]*x2[i,1]+beta2[2]*x2[i,2]+alpha[2]*x2[i,3]+s[g[i],2] } # distribution of the random effects for ( k in 1:p) { s[i,1:2]~dmnorm(mu0[],tau3[,]) } for ( k in 1:2) { mu0[k]~dnorm(0,0.001) beta1[k]~dnorm(0,0.001) beta2[k]~dnorm(0,0.001) } # correlated Prior distribution on the treatment effect alpha[1:2]~dmnorm(betamu1[],Sigma1[1:2,1:2]) Sigma1[1:2,1:2]<-inverse(omega[,]) tau1[1:2,1:2]~dwish( RT1[,] , 2 ) tau3[1:2,1:2]~dwish( RT3[,] , 2 ) omega1[1:2,1:2]<-inverse(tau1[,]) omega3[1:2,1:2]<-inverse(tau3[,]) omega[1,1]<-S2BR omega[2,2]<-S2BT omega[1,2]<-rho*sqrt(S2BT*S2BR) omega[2,1]<-omega[1,2] S2BR<-1/TAUBR S2BT<-1/TAUBT TAUBR<-0.055 TAUBT<-0.055 rho~dunif(0,1) RT1[1,1]<-1 RT1[1,2]<-0 RT1[2,1]<-0 RT1[2,2]<-1 RT3[1,1]<-1 RT3[1,2]<-0 RT3[2,1]<-0 RT3[2,2]<-1 # ABE criteria abe1<-2*alpha[1] abe2<-2*alpha[2] #posterior probability of criteria PAB1<-step(abe1-log(0.8))-step(abe1-log(1.25)) PAB2<-step(abe2-log(0.8))-step(abe2-log(1.25)) # joint posterior probability PAB<-PAB1*PAB2 } Data list(N=36,p=18, x1=structure(.Data=c( 0,0,0, 0,0,0, �. �. ), .Dim=c(36,3)), x2=structure(.Data=c( 0,0,0, 0,0,0, �. �. ), .Dim=c(36,3)), z=structure(.Data=c( 2.52,1.04, 8.87,3.80, � � ),.Dim=c(36,2)))