cvSingleModalityAdmm {HDMAADMM}

Cross Validation for High-dimensional Single Mediation Models

Description

Cross Validation for High-dimensional Single Mediation Models

Usage

cvSingleModalityAdmm(
  X,
  Y,
  M1,
  numFolds = 10,
  typeMeasure = "rmse",
  lambda1a,
  lambda1b,
  lambda1g,
  lambda2a,
  lambda2b,
  rho = 1,
  penalty = "ElasticNet",
  penaltyParameterList = list(),
  SIS = FALSE,
  SISThreshold = 2,
  maxIter = 3000,
  tol = 1e-04,
  verbose = FALSE,
  debug = FALSE
)

Arguments

X

The matrix of independent variables (exposure/treatment/group).

Y

The vector of dependent variable (outcome response).

M1

The single-modality mediator.

numFolds

The number of folds. The default is 10. Although nfolds can be as large as the sample size (leave-one-out CV), it is not recommended for large datasets. Smallest value allowable is nfolds=3.

typeMeasure

Default is "rmse".

rho, lambda1g, lambda1a, lambda1b, lambda2a, lambda2b, penaltyParameterList

Allow to put sequences for each parameter. Please refer to the function, singleModalityAdmm for the details.

penalty, SIS, SISThreshold, maxIter, tol, verbose, debug

Please refer to the function, singleModalityAdmm.

Value

An cvSingleModalityAdmm object which is a matrix containing all the combinations of parameter sequences with an additional column called measure.

Examples

## Generate Empirical Data
simuData <- modalityMediationDataGen(seed = 20231201)

## Cross-Validation for ElasticNet penalty
cvElasticNetResults <- cvSingleModalityAdmm(
  X = simuData$MediData$X, Y = simuData$MediData$Y, M1 = simuData$MediData$M1,
  numFolds = 5, typeMeasure = "rmse",
  rho = c(0.9, 1, 1.1), lambda1a = c(0.1, 0.5, 1), lambda1b = c(0.1, 0.3),
  lambda1g = c(1, 2), lambda2a = c(0.5, 1), lambda2b = c(0.5, 1),
  penalty = "ElasticNet"
)

## Cross-Validation for Pathway Lasso penalty (lambda2a, lambda2b are not tuned.)
cvPathwayLassoResults <- cvSingleModalityAdmm(
  X = simuData$MediData$X, Y = simuData$MediData$Y, M1 = simuData$MediData$M1,
  numFolds = 5, typeMeasure = "rmse",
  rho = c(0.9, 1, 1.1), lambda1a = c(0.1, 0.5, 1), lambda1b = c(0.1, 0.3),
  lambda1g = c(1, 2), lambda2a = 1, lambda2b = 1,
  penalty = "PathwayLasso", penaltyParameterList = list(kappa = c(0.5, 1), nu = c(1, 2))
)