Learn R Programming

gets (version 0.40)

ISMatricesLoop: Internal function to loop Over Indicator Saturation Matrices with Block-Based GETS

Description

Loops through each IS matrix and performs block-based GETS model selection, using optional parallelisation.

Usage

ISMatricesLoop(blocks.is.list, ISmatrices, ratio.threshold, mXncol,
  parallel.options, i, ISblocks, y, y.n, mX, userEstArg, t.pval, wald.pval, 
  do.pet, arLjungB, archLjungB, normality.JarqueB, user.diagnostics, gofFunArg, 
  gof.method, mxkeep, include.gum, include.1cut, include.empty, max.paths, 
  turbo, tol, LAPACK, max.regs, print.searchinfo, clusterSpec, clusterVarlist, 
  blocks, max.block.size, mXnames, additional.block.search)

Value

A list with:

ISfinalmodel

Vector of selected indicator names.

estimations.counter

Updated estimation counter.

getsFun.counter

Updated function call counter.

ISblocks

Final block structure.

Arguments

blocks.is.list

Logical; if TRUE, assumes that ISblocks already contains a user-supplied block structure. If FALSE, blocks are generated internally.

ISmatrices

List of indicator saturation matrices, typically generated by create.ISmatrices, containing IIS, SIS, TIS, or UIS.

ratio.threshold

Numeric value used to determine the minimum ratio of observations to regressors in each block.

mXncol

Number of columns in the original regressor matrix mX, used to ensure feasible model sizes.

parallel.options

If not NULL, enables parallel processing with specified number of cores or a cluster object.

i

Index of the current IS matrix in ISmatrices being processed.

ISblocks

List of index blocks for each IS matrix. Can be partially pre-defined or built internally.

y

Response variable, a numeric vector or time-series.

y.n

Number of observations in the response variable.

mX

Matrix of regressors excluding indicator variables.

userEstArg

List specifying the user-defined estimator function and its arguments, passed to getsFun.

t.pval

Significance threshold for t-tests in the GETS model selection process.

wald.pval

Significance threshold for Wald tests in the GETS process.

do.pet

Logical; if TRUE, Parsimonious Encompassing Tests are performed during deletion steps.

arLjungB

List with AR order and significance level for the Ljung-Box test for autocorrelation in residuals.

archLjungB

List with ARCH order and significance level for testing heteroskedasticity in residuals.

normality.JarqueB

Significance level for the Jarque-Bera normality test. If NULL, the test is skipped.

user.diagnostics

List containing user-specified diagnostics function and p-value settings, passed to getsFun.

gofFunArg

List specifying the name and parameters of the goodness-of-fit function used.

gof.method

Character string specifying whether the GOF criterion is minimized ("min") or maximized ("max").

mxkeep

Integer indices of regressors that are forced to be retained in the model.

include.gum

Logical; if TRUE, the General Unrestricted Model is included as a terminal model.

include.1cut

Logical; if TRUE, the one-cut model is included as a terminal model.

include.empty

Logical; if TRUE, the empty model is included among terminal models.

max.paths

Maximum number of paths to explore in GETS model selection. If NULL, no limit is imposed.

turbo

Logical; if TRUE, avoids redundant path searches in GETS.

tol

Tolerance threshold for detecting multicollinearity among regressors.

LAPACK

Logical; if TRUE, uses LAPACK for QR decomposition in collinearity checks.

max.regs

Maximum number of regressions to perform per GETS path.

print.searchinfo

Logical; if TRUE, progress and diagnostics messages are printed during execution.

clusterSpec

Specification object passed to makeCluster to control parallel processing.

clusterVarlist

Character vector listing variable names to be exported to each cluster worker.

blocks

Integer specifying the number of blocks to divide the IS matrix into. If NULL, determined automatically.

max.block.size

Maximum number of indicators per block used during block construction.

mXnames

Character vector of regressor names in mX, used to identify retained terms.

additional.block.search

Default is FALSE. If isat() identifies (retains) too many indicators in the block search, a sensible model might not be estimable (e.g. due to a lack of available degrees of freedom). In this case, setting additional.block.search = TRUE carries out an additional block search using a Leave-one-Out method (e.g. for three blocks, indicators are assigned to block 1,2,3,1,2,3,1,2,3). This is an experimental feature.

See Also

See also isat and ISadditionalblocksearch