Optimise scale parameters in MixARgen models
em_est_dist.RdOptimise the scale parameters in MixAR models from class MixARgen. Internal function.
Arguments
- tau
conditional probabilities, an object of class "MixComp", see 'Details'.
- etk
component residuals, see 'Details'.
- parscore
the score function(s), see 'Details'.
- sigma
current values of the scale parameters, a numeric vector.
- nu
current values of the parameters. w.r.t. which optimisation is done.
- logpdf
the log of pdf as a function of the parameters.
Details
One or more of the error distributions of a MixAR model may have
parameters that are considered unknown. In that case
em_est_dist can be used to optimise with respect to them.
The representation of the error distributions in "MixARgen" models
carries all the necessary information about
parameters. em_est_dist works by extracting their current
values from logpdf, passes them to the optimisation function
(or equation solver) and stores the result back into logpdf.
em_est_dist is quite general, as long as logpdf is
prepared according to the conventions it expects (this is so if they
are valid elements of the dist slot of "MixARgen" objects).