Error in parsing model: base type mismatch in assignment

Modeling
1

Hi,

I am new to Stan and am trying to construct a hierarchical model as per the following requirements:

response Y ~ neg_binomial_2( mu , phi )

where, mu = a + b*log(X)

a and b are drawn from the hyper-distribution that has a mean defined by a linear relationship with third variable (see model file, covariate U). However, I am getting the following error when I run the model:

SYNTAX ERROR, MESSAGE(S) FROM PARSER:

base type mismatch in assignment; variable name = u_gamma_t, type = real; right-hand side type=row vector
  error in 'model21844a1e2888_c1d0a5d3f064fe151ed16c4649a622e1' at line 41, column 15
  -------------------------------------------------
    39: a_trap ~ normal(0,sigma_trap);
    40: for (j in 1:J)
    41: u_gamma_t[j] = (U[j]*gamma);
                      ^
    42: beta ~ multi_normal(u_gamma_t, Sigma_beta);
  -------------------------------------------------

PARSER EXPECTED: <expression assignable to left-hand side>
Error in stanc(file = file, model_code = model_code, model_name = model_name,  : 
  failed to parse Stan model 'c1d0a5d3f064fe151ed16c4649a622e1' due to the above error.

There might be something very basic in the coding that I am missing. I’ve pasted my code below. Any help will be much appreciated!

Thank you for your time.

sts.ndd = "
data{
int<lower=1> N; // rows of observations
int<lower=0> Y[N]; // observed response
int<lower=1> K; //no.of predictors
int<lower=1> J; //no.of species
int<lower=1> L; // num group predictors
matrix[N,K] X; // individual predictors/seed numbers
matrix[J,L] U; // group predictors
int<lower=1, upper=J> jj[N]; // group for species	
int<lower=1> y; 						//number of  years
int<lower=1,upper=y> yy[N]; // group for year
int<lower=1> T; 						//number of  traps
int<lower=1,upper=T> tt[N]; // group for trap
}

parameters{
corr_matrix[K] Omega; //  correlation matrix
vector<lower=0>[K] tau; //  scale
vector[K] beta[J]; // indiv coeffs per species
matrix[L,K] gamma; // group coeffs 
vector[y] a_year;  					// year intercepts
real<lower=0> sigma_year;   // year sd
vector[T] a_trap;  					// trap intercepts
real<lower=0> sigma_trap;   // trap sd
}

model{
vector[N] x_beta_jj; // mean expectation
vector[K] u_gamma_t; //
matrix[K,K] Sigma_beta;
Sigma_beta = quad_form_diag(Omega,tau);
tau ~ cauchy(0,5);
Omega ~ lkj_corr(2.0); // prior covariance
to_vector(gamma) ~ normal(0,1);

// random effect priors
a_year ~ normal(0, sigma_year);
a_trap ~ normal(0,sigma_trap);
for (j in 1:J)
u_gamma_t[j] = (U[j]*gamma);
beta ~ multi_normal(u_gamma_t, Sigma_beta);

phi ~ gamma(0.01, 0.01); 

// Likelihood
for (n in 1:N)
x_beta_jj[n] = (X[n] * beta[jj[n]] + a_year[yy[n]] + a_trap[tt[n]]); //log location
sdl ~ neg_binomial_2_log(x_beta_jj, phi);
}

"

Best,
Meghna

2

I think you meant to declare vector[K] u_gamma_t[J]; in your model block.

3

Hi Ben,

Thanks for your response. I had initially declared it as you mentioned, but
was getting the same error. Which is why I had changed to the current
declaration.

Might there be an issue with how the data is provided? U is currently a
on-column matrix. Should I provide it as a column vector instead?

Thanks,
Meghna

4

Yes

5

Thanks Ben.

It appears that declaring row_vector[K] u_gamma_t[J]; works. The model did compile but I got the following runtime error:

   Error in new_CppObject_xp(fields$.module, fields$.pointer, ...) : 
  variable does not exist; processing stage=data initialization; variable name=yy; base type=int
In addition: Warning messages:
1: In FUN(X[[i]], ...) : data with name yy is not numeric and not used
2: In FUN(X[[i]], ...) : data with name tt is not numeric and not used
trying deprecated constructor; please alert package maintainer
Error in new_CppObject_xp(fields$.module, fields$.pointer, ...) : 
  no valid constructor available for the argument list
failed to create the sampler; sampling not done

So, as I understand, the grouping variables (in this case yy and tt) need to given as integers to Stan, not as factors. Is that right?

I don’t understand the error about deprecated constructor. I formulated my model based on instructions in the latest vignette. Apologies if I have missed something here.

6

Yes

Don’t worry about that.

7

Providing yy and tt as integers appears to be working.

Thank you very much for your help Ben! Appreciate the quick feedback.

8

Dear stan community. I am having a similar issue and have tried the solutions given here to no avail. I have therefore posted it as a separate question at SYNTAX ERROR, MESSAGE(S) FROM PARSER: Variable definition base type mismatch, variable declared as base type: vector variable definition has base.

Thank you in advance for any help.