mixAR models with Gaussian noise components

Class "MixARGaussian" represents MixAR models with Gaussian noise components.

Objects from the Class

Objects can be created by calls of the form new("MixARGaussian", ...), giving the elements of the model as named arguments, see the examples below. All elements of the model, except arcoef, are simple numeric vectors. From version 0.19-15 of package MixAR it is possible to create objects using MixARGaussian(...). The two forms are completely equivalent.

arcoef contains the AR coefficients, one numeric vector for each mixture component. It can be given as a "raggedCoef" object or as a list of numeric vectors.

To input a model with seasonal AR coefficients, argument passed to arcoef can be passed as a raggedCoefS object, or as a list of three elements. For the latter, seasonality s must be explicitly indicated. AR coefficients can be given as list or matrix within the main list (one for main AR coefficients, named a, and one for seasonal AR coefficients, as). Each row of a input matrix/element of the list denotes one component of the mixture. If not named, initialisation takes the first passed element to be a and the second to be as.

The AR order of the model is inferred from arcoef argument. If argument order is given, it is checked for consistency with arcoef. The shift slot defaults to a vector of zeroes and the scale slot to a vector of ones.

The distribution of the noise components is standard Gaussian, N(0,1).

Slots

All slots except arcoef are numeric vectors of length equal to the number of components in the model.

prob:: probabilities of the mixture components
order:: AR orders of the components
shift:: the shift (intercept) terms of the AR components
scale:: the standard deviations of the noise terms of the AR components
arcoef:: The AR components, object of class "raggedCoef"

Extends

Class "MixAR", directly.

Methods

mix_cdf: signature(model = "MixARGaussian", x = "numeric", index = "numeric", xcond = "missing"): ...
mix_cdf: signature(model = "MixARGaussian", x = "numeric", index = "missing", xcond = "numeric"): ...
fit_mixAR: signature(x = "ANY", model = "MixARGaussian", init = "MixAR"): ...
get_edist: signature(model = "MixARGaussian"): ...
mix_cdf: signature(model = "MixARGaussian", x = "missing", index = "missing", xcond = "numeric"): ...
mix_pdf: signature(model = "MixARGaussian", x = "missing", index = "missing", xcond = "numeric"): ...
mix_pdf: signature(model = "MixARGaussian", x = "numeric", index = "missing", xcond = "numeric"): ...
mix_pdf: signature(model = "MixARGaussian", x = "numeric", index = "numeric", xcond = "missing"): ...
noise_dist: signature(model = "MixARGaussian"): ...
noise_rand: signature(model = "MixARGaussian"): ...

Author

Georgi N. Boshnakov

Examples

showClass("MixARGaussian")
#> Class "MixARGaussian" [package "mixAR"]
#> 
#> Slots:
#>                                                              
#> Name:        prob      order      shift      scale     arcoef
#> Class:    numeric    numeric    numeric    numeric raggedCoef
#> 
#> Extends: "MixAR"

## load ibm data from BJ
## data(ibmclose, package = "fma")

## compute a predictive density, assuming exampleModels$WL_ibm model
## for the first date after the end of the data
pdf1 <- mix_pdf(exampleModels$WL_ibm, xcond = as.numeric(fma::ibmclose))
#> Registered S3 method overwritten by 'quantmod':
#>   method            from
#>   as.zoo.data.frame zoo 

## plot the predictive density
## (cdf is used to determine limits on the x-axis)
cdf1 <- mix_cdf(exampleModels$WL_ibm, xcond = as.numeric(fma::ibmclose))
gbutils::plotpdf(pdf1, cdf = cdf1, lq = 0.001, uq = 0.999)


## compute lower 5% quantile of cdf1
gbutils::cdf2quantile(0.05, cdf = cdf1)
#> [1] 347.7531