Estimating parameters of a SEIR model

Modeling
, ,
27

It’s neg_binomial_2, not neg_binomial_5 is the thing:

...
model {
  //priors
  phi ~ normal(0, 1); 
  pe ~ normal(0, 1); 
  pie ~ normal(0, 1); 
  pee ~ normal(0, 1); 
  beta ~ normal(0.5, 0.9); 
  mu ~ normal(0.2, 0.55); 
  betam ~ normal(0, 1);  
  tau ~ normal(1, 20); 
  alpha ~ normal(0.1, 0.99); 
  Ts ~ normal(3, 20); 
  Ta ~ normal(1, 20); 
  Tm ~ normal(1, 20); 
  Ti ~ normal(0.5, 8);
  phi_inv ~ exponential(5);
  
  //sampling distribution
  //col(matrix x, int n) - The n-th column of matrix x. Here the number of infected people 
  cases ~ neg_binomial_2(col(to_matrix(y), 5), phii);
}
generated quantities {
  real R0 = beta * Ta * (alpha * Ts * (1 - phi) / Ta + mu * (1 - alpha));
  real pred_cases[n_days];
  pred_cases = neg_binomial_2_rng(col(to_matrix(y), 5), phii);
}

With this the model compiles for me.

I feel like your rstan is not working correctly.

You shouldn’t be getting error messages like this:

Error in close.connection(zz)

You should be getting error messages like:

SYNTAX ERROR, MESSAGE(S) FROM PARSER:
Probability function must end in _lpdf or _lpmf. Found distribution family = neg_binomial_5 with no corresponding probability function neg_binomial_5_lpdf, neg_binomial_5_lpmf, or neg_binomial_5_log
 error in 'model66865f3d2989_sir_negbin' at line 113, column 53
  -------------------------------------------------
   111:   //sampling distribution
   112:   //col(matrix x, int n) - The n-th column of matrix x. Here the number of infected people 
   113:   cases ~ neg_binomial_5(col(to_matrix(y), 5), phii);
                                                            ^
   114: }
  -------------------------------------------------

Error in stanc(file = file, model_code = model_code, model_name = model_name,  : 
  failed to parse Stan model 'sir_negbin' due to the above error.

If you can, maybe try installing cmdstanr and switch to it? It is an alternative interface to Stan for R: https://mc-stan.org/cmdstanr/

If nothing else, maybe upgrade to R 4.0 and see if things start working? Only do that if you can risk other packages breaking though: R 4.0, rstan, and you

What operating system are you using? Is this Catalina? Or something else?

2 Likes