em_rinit: Gaussian EM-step with random initialisation

In mixAR: Mixture Autoregressive Models

Gaussian EM-step with random initialisation

Description

Gaussian EM-step with random initialisation.

Usage

em_rinit(y, order, partempl)

etk2tau(etk)

Arguments

y	time series.
order	MixAR order, vector of length the number of components.
partempl	parameter template, a list containing one element for each mixture component, see randomArCoefficients.
etk	MixAR component residuals, a matrix.

Details

em_rinit generates random MAR residuals, performs a non-distributional E-step, and a Gaussian M-step.

etk2tau estimates tau from component residuals only. Note that this is unlike em_tau, which also needs the noise pdf's, as well as estimates of the mixture probabilities.

em_rinit uses etk2tau to start the EM algorithm.

Value

for em_rinit, an object from class "MixARGaussian"

for etk2tau, a matrix representing tau (i-th row contains probabilities corresponding to the i-th observation)

Author(s)

Georgi N. Boshnakov

mixAR documentation built on May 3, 2022, 5:08 p.m.