Error : Rejecting initial value when parameter is ~ uniform(0, 1);

[Nikolay Evdokimov]

Hello, I want to realize simple idea  - to simulate poisson process with following code in Matlab:

>> lambda = 0.2;

int_times = (1/lambda)*-log(rand(10,1))

int_times =

    2.9671

    0.2108

    3.5610

    7.4034

   10.9117

    9.0102

    0.5532

    7.1497

    2.0342

    6.9926

So i try to make a model in Stan and run it.

model_dat <- list(rowsN = 100, creationsN=5, creationsLambda = 0.2)

fit <- stan(file = '.R/StockModels/Agents1.stan',data = model_dat,  iter = 100, chains = 1)

-----------
data {

  int<lower=0> rowsN; 

  int<lower=0> creationsN; 

  real creationsLambda;

}

parameters {

  vector[creationsN] interarrivalTime_U;

}

transformed parameters {

    vector[creationsN] interarrivalTime;

    vector[creationsN] arrivalTime;

    interarrivalTime <- 1/creationsLambda*-log(interarrivalTime_U);

    arrivalTime <- cumulative_sum(interarrivalTime);

}

model {

  interarrivalTime_U ~ uniform(0, 1);

}

But I get following error:

SAMPLING FOR MODEL 'Agents1' NOW (CHAIN 1).

  [1] "Error : Rejecting initial value:"                                                                     

  [2] "  Log probability evaluates to log(0), i.e. negative infinity."                                       

  [3] "  Stan can't start sampling from this initial value."                                                 

  [4] "Rejecting initial value:" 

What am I doing wrong here?

[Ben Goodrich]

On Tuesday, April 5, 2016 at 1:46:27 PM UTC-4, Nikolay Evdokimov wrote:

What am I doing wrong here?

Not constraining the parameter to the (0,1) interval, i.e.

parameters {
  vector<lower=0,upper=1>[creationsN] interarrivalTime_U;
}

at which point, having

interarrivalTime_U ~ uniform(0,1);


in the model block is redundant.

Ben