ForeCA (version 0.2.6)

complete-controls: Completes several control settings

Description

Completes algorithm, entropy, and spectrum control lists.

Usage

complete_algorithm_control(
  algorithm.control = list(max.iter = 50, num.starts = 10, tol = 0.001, type = "EM")
)
complete_entropy_control(
  entropy.control = list(base = NULL, method = "MLE", prior.probs = NULL, prior.weight
    = 0.001, threshold = 0),
  num.outcomes
)
complete_spectrum_control(
  spectrum.control = list(kernel = NULL, method = c("wosa", "direct", "multitaper",
    "mvspec", "ar", "pgram"), smoothing = FALSE)
)

Arguments

algorithm.control

list; control parameters for any iterative ForeCA algorithm.

entropy.control

list; control settings for entropy estimation.

num.outcomes

positive integer; number of outcomes for the discrete probability distribution. Must be specified (no default value).

spectrum.control

list; control settings for spectrum estimation.

Value

A list with fully specified algorithm, entropy, or spectrum controls. Default values are only added if the input {spectrum,entropy,algorithm}.control list does not already set this value.

complete_algorithm_control returns a list containing:

max.iter

maximum number of iterations; default: 50.

num.starts

number of random starts to avoid local optima; default: 10.

tol

tolerance for when convergence is reached in any iterative ForeCA algorithm; default: 1e-03.

type

string; type of algorithm. Default: 'EM'.

complete_entropy_control returns a list with:

base

logarithm base for the entropy.

method

string; method to estimate entropy; default: "MLE".

prior.probs

prior distribution; default: uniform rep(1 / num.outcomes, num.outcomes).

prior.weight

weight of the prior distribution; default: 1e-3.

threshold

non-negative float; set probabilities below threshold to zero; default: 0.

complete_spectrum_control returns a list containing:

kernel

R function; function to weigh each Fourier frequency \(\lambda\); default: NULL (no re-weighting).

method

string; method to estimate the spectrum; default: 'wosa' if sapa is installed, 'mvspec' if only astsa is installed, and 'pgram' if neither is installed.

smoothing

logical; default: FALSE.

Available methods for spectrum estimation are (alphabetical order)

"ar"

autoregressive spectrum fit via spec.ar; only for univariate time series.

"direct"

raw periodogram using SDF.

"multitaper"

tapering the periodogram using SDF.

"mvspec"

smoothed estimate using mvspec; many tuning parameters are available -- they can be passed as additional arguments (...) to mvspectrum.

"pgram"

uses mvpgram; is the same as the 'direct' method, but does not rely on the SDF package.

"wosa"

Welch overlapping segment averaging (WOSA) using SDF.

Setting smoothing = TRUE will smooth the estimated spectrum (again); this option is only available for univariate time series/spectra.

See Also

mvspectrum, discrete_entropy, continuous_entropy