Memory-efficient FFBS that avoids creating n^2 x n^2 matrices.
Supports rectangular (bipartite) Theta via row-covariance approximation.
Usage
ffbs_theta_struct_5arg_cpp(Z, mu, A_array, B_array, sigma2)
Arguments
- Z
Observations (n_row x n_col x T array as cube)
- mu
Baseline mean (n_row x n_col matrix)
- A_array
Time-varying A matrices (n_row x n_row x T)
- B_array
Time-varying B matrices (n_col x n_col x T)
- sigma2
Innovation variance (scalar)
Value
Sampled Theta array (n_row x n_col x T)