Issues with OpenBUGS and switch bugs codes to Stan

Interfaces
, ,
1

Hello! I am new to Stan and I am trying to switch the bug’s code of my spatial survival model to Stan. In OpenBUGS, my model is syntactically correct, data loaded, model compiled, initial values generated and model initialized. But, during the updating time of the model (model is updating), I encounter the following error in OpenBUGS:
“Sorry, something went wrong in procedure Stack. Value in module GraphStack”.
To tackle this problem, I am trying to switch the following bug’s code to Stan. Can anyone help me?

"# bug's code
model {
for (k in 1:q) {
for (j in 1:ng) {
landa[k,j] <- kernel [k,j] * exp(x[j])
}
z[k]<- sum(landa[k,])*p[k]
a[k]<- z[k]/sum(z[])
for (i in 1:N) {
#temp: (logt-beta0-beta*x-frailty) / sigma1
temp[k,i]<- (log(time[i]+0.01)-beta0-beta [1]*sex [i]-beta [2]*prison [i]-beta [3]*(age[i]-40)/20-beta [4]*smear.diag[i]-beta[5]*history[i]-W[county[i]]) / sigma
#survival and density function of extreme value distribution
s0[k,i]<-exp(-exp(temp[k,i]))
f0[k,i]<-s0[k,i]*exp(temp[k,i])
#loglikelihood function
L[k,i]<- censor.status[i]*log(f0[k,i])-censor.status[i]* log(sigma)-censor.status[i]*log(time[i]+0.01)+log(s0[k,i])+log(a[k])-censor.status[i]*log(s0[k,i])
zeros[k,i]<-0
zeros[k,i] ~dloglik(L[k,i])
}
}
#prior
inversesigma~ dgamma (1,2)
sigma<- 1/inversesigma
beta0~dnorm(0,0.001)
for(m in 1:5) {beta[m]~ dnorm(0,0.001)}
for(j in 1:ng) {x[j]~dnorm(0,0.04)}
for(i in 1:14) {W[i]~dnorm(0,0.001)}
}
#Initial Value
list(beta0=0,inversesigma=0.10,beta=c(0,0,0,0,0),x=c(0,0,0,0,0,0,0,0))
#Data
list(kernel =structure(.Data = c( 0.5158, 0.5853, 0.6622, 0.7440, 0.5257, 0.5994, 0.6833, 0.7787, 0.7102, 0.7641, 0.7793, 0.7452, 0.7686, 0.8550, 0.8872, 0.8216, 0.9454, 0.8349, 0.7330,0.6431, 0.8957, 0.8156, 0.7227, 0.6367, 0.8717, 0.8607, 0.7817, 0.6941, 0.9494, 0.9234, 0.8098, 0.7102,0.5822, 0.6598, 0.7436, 0.8250, 0.5936, 0.6769, 0.7717, 0.8798, 0.6648,0.7547, 0.8505, 0.9158, 0.6696, 0.7627, 0.8669, 0.9571, 0.9309, 0.9412, 0.8256, 0.7240, 0.8666, 0.8711, 0.7959, 0.7076,0.4144,0.4629,0.5131,0.5622,0.4393,0.4951,0.5553,0.6175,0.3315,0.3694,0.4091,0.4492,0.3537,0.3971,0.4437,0.4924,0.4767,0.5392,0.6073,0.6786,0.4914,0.5594,0.6362,0.7220,0.4178,0.4711,0.5288,0.5892,0.4350,0.4939,0.5596,0.6318,0.7955,0.9026,0.9452,0.8401,0.7850,0.8785,0.9074,0.8259,0.7840,0.7758,0.7242,0.6555,0.8896,0.8715,0.7849,0.6946,0.5022,0.5515,0.5957,0.6276,0.5424,0.6042,0.6636,0.7105), .Dim=c(14, 8)), q=14, ng=8,p=c(0.92,0.87,0.86,0.86,0.88,0.87,0.85,0.94,0.92,0.87,0.82,0.83,0.80,0.91))
list(N=2490)
county[] sex[]  prison[] age[] smear.diag[] history[]  censor.status[] time[]

Best regards,
Eisa

2

There’s a section in the Stan user guide on transitioning from BUGS: 25 Transitioning from BUGS | Stan User’s Guide

The way to do this is in pieces. For instance, start by just converting your priors from BUGS to Stan. Make sure you’re able to run your Stan program and analyze the results. Once you have that, code up the Stan data block and make sure you’re able to read in all your data into your Stan program. After you can read all the data in, start adding your likelihood terms (in small pieces, if possible, testing that the model runs along the way)

1 Like