Information message when using function `multi_normal_lpdf`

Modeling
,
1

Hi all,

I got some information messages when using the multi_normal_lpdf function as the following,

Informational Message: The current Metropolis proposal is about to be rejected because of the following issue:
Chain 4 Exception: multi_normal_lpdf: LDLT_Factor of covariance parameter is not positive definite.  last conditional variance is nan.

,

Exception: Exception: multi_normal_lpdf: Covariance matrix is not symmetric.

, or

Exception: Exception: quad_form_sym: A is not symmetric. A[1,2] = nan, but A[2,1] = nan

I tried to tune the step_size parameter to .01 to fix the problem, and it can be fixed to some extent. But I still wonder what causes the problem.

The equation that I pass to the multi_normal_lpdf function is as follows,

functions{
  real mmn(real mu_intercept, real mu_slope, vector time_seg, 
          matrix Sigma, real sigma_e, vector y_seg){
    vector[rows(time_seg)] mu_seg;
    matrix[rows(time_seg),2] Z_j;
    matrix[rows(time_seg),rows(time_seg)]Cov_j;
    mu_seg = mu_intercept + (mu_slope)* time_seg;
    Z_j = append_col(rep_vector(1.0, rows(time_seg)), time_seg);
    Cov_j = quad_form_sym(Sigma, Z_j') + diag_matrix(rep_vector(sigma_e^2, rows(time_seg)));
    return multi_normal_lpdf (y_seg| mu_seg, Cov_j);
  }
}

Can anyone help me? I appreciate it very much.

Best wishes,
Doria

2

How do you declare/construct the matrix that you pass to this function as Sigma?