Struct DynamicFactorModel 

pub struct DynamicFactorModel {
    pub loadings: Array2<f64>,
    pub ar_matrices: Vec<Array2<f64>>,
    pub factors: Array2<f64>,
    pub r: Array2<f64>,
    pub q: Array2<f64>,
    pub n_factors: usize,
    pub n_vars: usize,
    pub n_lags: usize,
    pub log_lik_history: Vec<f64>,
}

Fitted Dynamic Factor Model.

All fields correspond to the last EM iterate (or the converged solution when max_iter is large enough).

Fields

loadings: Array2<f64>

Factor loading matrix Λ of shape (n_vars, n_factors).

ar_matrices: Vec<Array2<f64>>

VAR coefficient matrices [A_1, ..., A_p], each (n_factors, n_factors).

factors: Array2<f64>

Smoothed factors, shape (T, n_factors).

r: Array2<f64>

Idiosyncratic covariance matrix R, shape (n_vars, n_vars) (diagonal).

q: Array2<f64>

Factor innovation covariance Q, shape (n_factors, n_factors).

n_factors: usize

Number of latent factors.

n_vars: usize

Number of observed variables.

n_lags: usize

Number of VAR lags.

log_lik_history: Vec<f64>

EM log-likelihood trajectory (one value per iteration).

Implementations

impl DynamicFactorModel

pub fn extract_factors(&self) -> Array2<f64>

Return the smoothed factor matrix (rows = time, cols = factors).

This is a convenience alias for the factors field.

pub fn forecast_one_step(&self) -> StatsResult<Array1<f64>>

One-step-ahead forecast of the observations at time T+1, given the last smoothed factor state.

Returns a K-vector X̂_{T+1} = Λ * E[F_{T+1} | X_{1:T}].

Trait Implementations

impl Clone for DynamicFactorModel

fn clone(&self) -> DynamicFactorModel

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for DynamicFactorModel

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.