lik_params {mixAR} | R Documentation |
Vector of parameters of a MixAR model
Description
Give a numeric vector containing non-redundant parameters of a MixAR model in a form suitable for use by optimisation routines. The methods defined in package mixAR for this generic function are described here.
Usage
lik_params(model)
Arguments
model |
a MixAR model. |
Details
lik_params
gives the parameters of a MixAR model as a numeric
vector.
This is a generic function. Parameters common to all MixAR models are arranged as described below. There are no other parameters when the error distributions do not contain parameters of their own. Methods for sub-classes with additional parameters should append them after the common parameters.
If k
is the number of components and \pi_i
is the
probability associated with the i
th component, then the
parameters are put in a vector as follows:
-
component probabilities,
\pi_1,\ldots,\pi_{k-1}
, (note:\pi_{k}
is not included) -
scales,
\sigma_1,\ldots,\sigma_{k}
, -
shifts,
\mu_1,\ldots,\mu_{k}
, -
AR coefficients of the 1st component,
-
AR coefficients of the 2nd component,
-
...
-
AR coefficients of the
k
th component.
Value
A numeric vector containing all parameters except the probability associated with the last component.
Methods
signature(model = "MixAR")
signature(model = "MixARgen")
Note
The probability associated with the k
th component is omitted
as it is redundant. This makes it possible to try unconstrained
optimisation though it is not likely to give useful results since
there are other restrictions on the probabilities.
Author(s)
Georgi N. Boshnakov