Problem with using rstan

ABUJARAD · September 10, 2018, 3:52pm

library(rstan)
model_code1="
functions{
//defined survival
vector log_s(vector t, real shape1, vector shape2){
vector[num_elements(t)] log_s;
for(i in 1:num_elements(t)){
log_s[i]=log(1-(1-(1+t[i])^(-shape2[i]) )^(shape1));

//log_s[i]=log(((1+t[i])^(scale[i]*shape) - ((1+t[i])^(scale[i]) -1)^shape )/((1 // +t[i])^(scale[i]*shape)));
}
return log_s;
}

//define log_ft
vector log_ft(vector t, real shape1, vector shape2){
vector[num_elements(t)] log_ft;
for(i in 1:num_elements(t)){
log_ft[i]=log((shape1 * shape2[i]) * (1-(1+t[i])^(shape2[i]))^(shape1-1) * (1+t[i])^-(shape2[i]+1));
}
return log_ft;
}
//define log hazard
vector log_h(vector t, real shape1, vector shape2){
vector[num_elements(t)] log_h;
vector[num_elements(t)] logft;
vector[num_elements(t)] logs;
logft=log_ft(t,shape1,shape2);
logs=log_s(t,shape1,shape2);
log_h=logft-logs;
return log_h;
}

//define the sampling distribution
real surv_expP_lpdf(vector t, vector d, real shape1 , vector shape2){
vector[num_elements(t)] log_lik;
real prob;
log_lik=d .* log_h(t,shape1,shape2)+log_s(t,shape1,shape2);
prob=sum(log_lik);
return prob;
}
}
//data block

data {
int N; // number of observations
vector<lower=0>[N] y; // observed times
vector<lower=0,upper=1>[N] censor;//censoring indicator (1=observed, 0=censored)
int M; // number of covariates
matrix[N, M] x; // matrix of covariates (with N rows and M columns)
}

parameters {
vector[M] beta; // Coefficients in the linear predictor (including intercept)
real<lower=0> shape1; // shape parameter

}

transformed parameters {
vector[N] linpred;
vector[N] shape2;
linpred = x * beta;
for (i in 1:N) {
shape2[i] = exp(linpred[i]);
}
}

model {
shape1 ~ uniform(0, 10);
beta ~ normal(0, 1);
y ~ surv_expP (censor, shape1, shape2);
}
generated quantities{
real dev;
dev=0;
dev=dev + (-2)*surv_expP_lpdf (y|censor,shape1,shape2);
}
"

data creation in R

y<-c(23,47,69,70,71,100,101,148,181,198,208,212,224,5,8,10,13,18,24,26,26,31,35,40,41,48,50,59,61,68,71,76,105,107,109,113,116,118,143,154,162,188,212,217,225)
x1<-c(rep(0,13), rep(1,32))
censor<-c(rep(1,3),rep(0,4),rep(1,2),rep(0,4),rep(1,18),rep(0,4),1,0,1,1,rep(0,6))
x <- cbind(1,x1)
N = nrow(x)
M = ncol(x)
event=censor
dat <- list( y=y, x=x, event=event, N=N, M=M)
M1<-stan(model_code=model_code1,data=dat,init=list(list(beta=c(-0.6,0.5))),iter=1000,chains=1)

I find this error!!!

Log probability evaluates to log(0), i.e. negative infinity.
Stan can’t start sampling from this initial value.
Rejecting initial value:
Log probability evaluates to log(0), i.e. negative infinity.
Stan can’t start sampling from this initial value.

Initialization between (-2, 2) failed after 100 attempts.
Try specifying initial values, reducing ranges of constrained values, or reparameterizing the model.
[1] “Error in sampler$call_sampler(args_list[[i]]) : Initialization failed.”
[1] “error occurred during calling the sampler; sampling not done”

bgoodri · September 10, 2018, 4:03pm

The initialization error is presumably due to the fact that you declare shape1 to not have an upper bound but put zero density for all values greater than 10. So, start by getting rid of that uniform prior, which is a bad idea anyway. There may be other problems.

ABUJARAD · September 10, 2018, 4:13pm

Thanks for you
But I tried to use
shape1 ~ cauchy(0,5);
and
shape1~ normal (0,10);
i get the same problem

bgoodri · September 10, 2018, 4:45pm

Then try specifying init_r = 0.5 or some number less than 2 when you call Stan.

ABUJARAD · September 10, 2018, 4:48pm

i do it, i get the same error…

ABUJARAD · September 10, 2018, 4:52pm

and i use init="random" when i call Stan, but get same problem

bgoodri · September 10, 2018, 5:15pm

I would put

if (!is_finite(prob)) {
  print("prob is ", prob);
  // print more stuff
}

into the end of surv_expP_lpdf to figure out what goes wrong.

ABUJARAD · September 10, 2018, 5:39pm

Sorry but I did not understand your intention.

Bob_Carpenter · September 10, 2018, 5:44pm

Debug by print. Put print statements into your code to figure out where things become infinite. It’s the only way to track things down.

ABUJARAD · September 10, 2018, 5:49pm

Thanks Bob, but where can i but that code ?

bgoodri · September 10, 2018, 6:03pm

Before the return statement in your surv_expP_lpdf function.

ABUJARAD · September 10, 2018, 6:28pm

I run it but i find this
`> model_code1="

functions{
//defined survival
vector log_s(vector t, real shape1, vector shape2){
vector[num_elements(t)] log_s;
for(i in 1:num_elements(t)){
log_s[i]=log(1-(1-(1+t[i])^(-shape2[i]) )^(shape1));
//log_s[i]=log(((1+t[i])^(scale[i]*shape) - ((1+t[i])^(scale[i]) -1)^shape )/((1 // +t[i])^(scale[i]*shape)));
}
return log_s;
}
//define log_ft
vector log_ft(vector t, real shape1, vector shape2){
vector[num_elements(t)] log_ft;
for(i in 1:num_elements(t)){
log_ft[i]=log((shape1 * shape2[i]) * (1-(1+t[i])^(shape2[i]))^(shape1-1) * (1+t[i])^-(shape2[i]+1));
}
return log_ft;
}
//define log hazard
vector log_h(vector t, real shape1, vector shape2){
vector[num_elements(t)] log_h;
vector[num_elements(t)] logft;
vector[num_elements(t)] logs;
logft=log_ft(t,shape1,shape2);
logs=log_s(t,shape1,shape2);
log_h=logft-logs;
return log_h;
}
//define the sampling distribution
real surv_expP_lpdf(vector t, vector d, real shape1 , vector shape2){
vector[num_elements(t)] log_lik;
real prob;
log_lik=d .* log_h(t,shape1,shape2)+log_s(t,shape1,shape2);
prob=sum(log_lik);
if (!is_finite(prob)) { print(“prob is”, prob); // print more stuff }
Error: unexpected symbol in:
"prob=sum(log_lik);
if (!is_finite(prob)) { print(“prob”

return prob;
Error: unexpected symbol in “return prob”
}
Error: unexpected ‘}’ in “}”
}
Error: unexpected ‘}’ in “}”
//data block
Error: unexpected ‘/’ in “/”

data {
Error: unexpected ‘{’ in “data {”
int N; // number of observations
Error: unexpected symbol in “int N”
vector<lower=0>[N] y; // observed times
Error: unexpected ‘[’ in “vector<lower=0>[”
vector<lower=0,upper=1>[N] censor;//censoring indicator (1=observed, 0=censored)
Error: unexpected ‘,’ in “vector<lower=0,”
int M; // number of covariates
Error: unexpected symbol in “int M”
matrix[N, M] x; // matrix of covariates (with N rows and M columns)
Error: unexpected symbol in “matrix[N, M] x”
}
Error: unexpected ‘}’ in “}”

parameters {
Error: unexpected ‘{’ in “parameters {”
vector[M] beta; // Coefficients in the linear predictor (including intercept)
Error: unexpected symbol in “vector[M] beta”
real<lower=0> shape1; // shape parameter
Error: object ‘shape1’ not found

}
Error: unexpected ‘}’ in “}”

transformed parameters {
Error: unexpected symbol in “transformed parameters”
vector[N] linpred;
Error: unexpected symbol in “vector[N] linpred”
vector[N] shape2;
Error: unexpected symbol in “vector[N] shape2”
linpred = x * beta;
for (i in 1:N) {

shape2[i] = exp(linpred[i]);
}
Error: object ‘shape2’ not found

}
Error: unexpected ‘}’ in “}”

model {
Error: unexpected ‘{’ in “model {”
shape1 ~ normal(0, 2);
shape1 ~ normal(0, 2)
beta ~ normal(0, 2);
beta ~ normal(0, 2)
y ~ surv_expP (censor, shape1, shape2);
y ~ surv_expP(censor, shape1, shape2)
}
Error: unexpected ‘}’ in “}”
generated quantities{
Error: unexpected symbol in “generated quantities”
real dev;
Error: unexpected symbol in “real dev”
dev=0;
dev=dev + (-2)*surv_expP_lpdf (y|censor,shape1,shape2);
Error in surv_expP_lpdf(y | censor, shape1, shape2) :
could not find function “surv_expP_lpdf”

}
Error: unexpected ‘}’ in “}”
"`

bgoodri · September 10, 2018, 6:31pm

You commented out the closing }. Also, you should print more stuff in that block in order to figure out under what conditions it goes wrong.

ABUJARAD · September 10, 2018, 6:36pm

then the same
`> model_code1="

functions{
//defined survival
vector log_s(vector t, real shape1, vector shape2){
vector[num_elements(t)] log_s;
for(i in 1:num_elements(t)){
log_s[i]=log(1-(1-(1+t[i])^(-shape2[i]) )^(shape1));
//log_s[i]=log(((1+t[i])^(scale[i]*shape) - ((1+t[i])^(scale[i]) -1)^shape )/((1 // +t[i])^(scale[i]*shape)));
}
return log_s;
}
//define log_ft
vector log_ft(vector t, real shape1, vector shape2){
vector[num_elements(t)] log_ft;
for(i in 1:num_elements(t)){
log_ft[i]=log((shape1 * shape2[i]) * (1-(1+t[i])^(shape2[i]))^(shape1-1) * (1+t[i])^-(shape2[i]+1));
}
return log_ft;
}
//define log hazard
vector log_h(vector t, real shape1, vector shape2){
vector[num_elements(t)] log_h;
vector[num_elements(t)] logft;
vector[num_elements(t)] logs;
logft=log_ft(t,shape1,shape2);
logs=log_s(t,shape1,shape2);
log_h=logft-logs;
return log_h;
}
//define the sampling distribution
real surv_expP_lpdf(vector t, vector d, real shape1 , vector shape2){
vector[num_elements(t)] log_lik;
real prob;
log_lik=d .* log_h(t,shape1,shape2)+log_s(t,shape1,shape2);
prob=sum(log_lik);
if (!is_finite(prob)) { print(“prob is”, prob);} // print more stuff
Error: unexpected symbol in:
"prob=sum(log_lik);
if (!is_finite(prob)) { print(“prob”

return prob;
Error: unexpected symbol in “return prob”
}
Error: unexpected ‘}’ in “}”
}
Error: unexpected ‘}’ in “}”
//data block
Error: unexpected ‘/’ in “/”

data {
Error: unexpected ‘{’ in “data {”
int N; // number of observations
Error: unexpected symbol in “int N”
vector<lower=0>[N] y; // observed times
Error: unexpected ‘[’ in “vector<lower=0>[”
vector<lower=0,upper=1>[N] censor;//censoring indicator (1=observed, 0=censored)
Error: unexpected ‘,’ in “vector<lower=0,”
int M; // number of covariates
Error: unexpected symbol in “int M”
matrix[N, M] x; // matrix of covariates (with N rows and M columns)
Error: unexpected symbol in “matrix[N, M] x”
}
Error: unexpected ‘}’ in “}”

parameters {
Error: unexpected ‘{’ in “parameters {”
vector[M] beta; // Coefficients in the linear predictor (including intercept)
Error: unexpected symbol in “vector[M] beta”
real<lower=0> shape1; // shape parameter
Error: object ‘shape1’ not found

}
Error: unexpected ‘}’ in “}”

transformed parameters {
Error: unexpected symbol in “transformed parameters”
vector[N] linpred;
Error: unexpected symbol in “vector[N] linpred”
vector[N] shape2;
Error: unexpected symbol in “vector[N] shape2”
linpred = x * beta;
for (i in 1:N) {

shape2[i] = exp(linpred[i]);
}
Error: object ‘shape2’ not found

}
Error: unexpected ‘}’ in “}”

model {
Error: unexpected ‘{’ in “model {”
shape1 ~ normal(0, 2);
shape1 ~ normal(0, 2)
beta ~ normal(0, 2);
beta ~ normal(0, 2)
y ~ surv_expP (censor, shape1, shape2);
y ~ surv_expP(censor, shape1, shape2)
}
Error: unexpected ‘}’ in “}”
generated quantities{
Error: unexpected symbol in “generated quantities”
real dev;
Error: unexpected symbol in “real dev”
dev=0;
dev=dev + (-2)*surv_expP_lpdf (y|censor,shape1,shape2);
Error in surv_expP_lpdf(y | censor, shape1, shape2) :
could not find function “surv_expP_lpdf”

}
Error: unexpected ‘}’ in “}”
"

`

ABUJARAD · September 10, 2018, 7:16pm

I DONT know what is this error?

bgoodri · September 10, 2018, 7:57pm

I think it misencoded the quotation marks. Moreover, there actually isn’t an is_finite function in the Stan language, so you should use

if (is_inf(prob)) {
  print("prob is ", prob);
  // print more stuff
}

ABUJARAD · September 10, 2018, 8:04pm

this is i get

    `Error: unexpected symbol in:

"if (is_inf(prob)) {
print(“prob”

}// print more stuff
Error: unexpected ‘}’ in “}”
return prob;
Error: unexpected symbol in “return prob”
}
Error: unexpected ‘}’ in “}”
}
Error: unexpected ‘}’ in “}”
//data block
Error: unexpected ‘/’ in “/”

data {
Error: unexpected ‘{’ in “data {”
int N; // number of observations
Error: unexpected symbol in “int N”
vector<lower=0>[N] y; // observed times
Error: unexpected ‘[’ in “vector<lower=0>[”
vector<lower=0,upper=1>[N] censor;//censoring indicator (1=observed, 0=censored)
Error: unexpected ‘,’ in “vector<lower=0,”
int M; // number of covariates
Error: unexpected symbol in “int M”
matrix[N, M] x; // matrix of covariates (with N rows and M columns)
Error: unexpected symbol in “matrix[N, M] x”
}
Error: unexpected ‘}’ in “}”

parameters {
Error: unexpected ‘{’ in “parameters {”
vector[M] beta; // Coefficients in the linear predictor (including intercept)
Error: unexpected symbol in “vector[M] beta”
real<lower=0> shape1; // shape parameter
Error: object ‘shape1’ not found

}
Error: unexpected ‘}’ in “}”

transformed parameters {
Error: unexpected symbol in “transformed parameters”
vector[N] linpred;
Error: unexpected symbol in “vector[N] linpred”
vector[N] shape2;
Error: unexpected symbol in “vector[N] shape2”
linpred = x * beta;
for (i in 1:N) {

shape2[i] = exp(linpred[i]);
}
Error: object ‘shape2’ not found

}
Error: unexpected ‘}’ in “}”

model {
Error: unexpected ‘{’ in “model {”
shape1 ~ normal(0, 2);
shape1 ~ normal(0, 2)
beta ~ normal(0, 2);
beta ~ normal(0, 2)
y ~ surv_expP (censor, shape1, shape2);
y ~ surv_expP(censor, shape1, shape2)
}
Error: unexpected ‘}’ in “}”
generated quantities{
Error: unexpected symbol in “generated quantities”
real dev;
Error: unexpected symbol in “real dev”
dev=0;
dev=dev + (-2)*surv_expP_lpdf (y|censor,shape1,shape2);
Error in surv_expP_lpdf(y | censor, shape1, shape2) :
could not find function “surv_expP_lpdf”

}
Error: unexpected ‘}’ in “}”
"`

ABUJARAD · September 10, 2018, 8:05pm

if you can run it, and get my the solu ,
thanks for you

bgoodri · September 10, 2018, 8:16pm

Those quotation marks are not ASCII. If you copy-and-paste, you need to delete them and put in plain quotation marks.

ABUJARAD · September 10, 2018, 8:18pm

if you want i can send the file to you e mail to run it in your system

Topic		Replies	Views
Issues with running survival model in Stan Modeling rstan , techniques , fitting-issues	4	466	May 24, 2021
Survival analysis: initial values rejected Modeling	23	1628	August 11, 2018
Problem running RStan with simulated data in R RStan rstan , fitting-issues	26	1895	February 5, 2022
Model returning Log probability evaluates to log(0), i.e. negative infinity Modeling	3	348	October 7, 2021
Initialization and log probability problem RStan	1	422	April 8, 2020

Problem with using rstan

data creation in R

Related topics