update_pars: Perform one EM-step

In b-brune/tvRRR: Time-varying reduced-rank regression using state-space models

Description

Function that updates the parameters from the current output of the KF-function i.e. one updating step in the EM algorithm

Usage

update_pars(
  kf,
  p,
  d,
  t,
  q,
  tol = 0.001,
  maxit = 50,
  Omega_diagonal = FALSE,
  model = "A"
)

Arguments

kf	a Kalman filter object, output from eval_tvRRR() / filter_modelA() / filter_modelB() function
p	dimension of the target
d	latent dimension
t	length of the time series
q	dimension of the predictors
tol	tolerance for convergence in iterative parameter updating
Omega_diagonal	logical, indicates whether Omega is assumed to be diagonal

Value

List of updated parameters:

Omega
beta	(when fitting model A)
alpha	(when fitting model B)
Sigma
Gamma	(when considering external variables)
Q	value of the expected likelihood Q(Theta|Theta^j)

b-brune/tvRRR documentation built on Dec. 19, 2021, 6:37 a.m.