PruneForwardStepwise: Given multiple composite predictors, prune the predictors...

In ncats/MultiOmicsGraphPrediction: Network-Based Ensemble Prediction from IntLIM Predictors

Given multiple composite predictors, prune the predictors that are not needed.

Description

Given multiple composite predictors, prune the predictors that are not needed.

Usage

PruneForwardStepwise(
  pairs,
  targets,
  sources,
  modelResults,
  minCutoff,
  maxCutoff,
  useCutoff,
  weights,
  pred,
  verbose,
  mapping,
  trueVal,
  pruningMethod,
  previousModels,
  individualPerformance,
  tolerance,
  i,
  numToAdd,
  modelRetention,
  averaging,
  zeroOut = FALSE
)

Arguments

pairs	A list of pairs to include in the composite model.
targets	A list of targets to include in the composite model.
sources	A list of sources to include in the composite model.
modelResults	The ModelResults object that will be filled in during training, obtained from DoModelSetup().
minCutoff	Mininum cutoff for the prediction.
maxCutoff	Maximum cutoff for the prediction.
useCutoff	Whether or not to use the cutoff for prediction. Default is FALSE.
weights	The weights for each predictor, calculated using ComputeMetaFeatureWeights()
pred	The predicted values using each individual predictor.
verbose	Whether or not to print out each step.
mapping	A mapping from models to composite models for the next stage of pooling.
trueVal	The true values of the phenotype.
pruningMethod	The method to use for pruning. Right now, only "error.t.test" is valid.
previousModels	A list of the previous models that were consolidated.
individualPerformance	The score (using the pruning method) for each individual component of the model.
tolerance	Tolerance factor when computing equality of two numeric values.
i	Model index
numToAdd	Number of models to remove at once when pruning.
modelRetention	Strategy for model retention. "stringent" (the default) retains only models that improve the prediction score. "lenient" also retains models that neither improve nor reduce the prediction score.
averaging	If TRUE, then averaging is used to combine predictors rather than retaining the same functional form for both the input and the output.
zeroOut	This parameter zeros out predictors outside of the allowed range.

Value

A ModelIDSet with the retained models.

ncats/MultiOmicsGraphPrediction documentation built on Aug. 23, 2023, 9:19 a.m.