Calculates estimates of scale parameters of MixAR models from conditional probabilities and mixture `residuals'. Used in EM algorithm.
tauetk2sigmahat(tau, etk)em_est_sigma(tau, etk, Fscore, sigma,
dontfix = rep(TRUE, length(sigma)), compwise = FALSE)
The new values of the scale parameters, a numeric vector
the conditional probabilities for the groups, a "MixComp"
object.
component "residuals", MixComp object(?).
the score function(s) of the noise distributions.
current values of the scale parameters.
if TRUE solve the equations component-wise, see
'Details.
a logical vector containig TRUE in the positions
of elements of sigma that are to be estimated.
tauetk2sigmahat calculates estimates of the scale parameters
for a MixAR time series with Gaussian components. There is an explicit
formula in that case.
em_est_sigma calculates estimates of the scale parameters in
the general case. The non-linear equations are solved using functions
from package BB. The equations for the components can often be
solved independently. When that is the case, compwise may speed
things up a little.