data { 
  int<lower=1> N; // number of data points
  vector[N] z;    // observations

} 
parameters {
  real<lower=0> mu; 
} 
model { 
  mu ~ normal(0,5);
  
  for(n in 1:N)
    target += normal_lpdf(z[n] | mu, 1) - normal_lccdf(1.96 | mu, 1);
  
}