RunCrossValidation: Runs the cross-validation end-to-end using the following...

In IntLIM: Integration of Omics Data Using Linear Modeling

Runs the cross-validation end-to-end using the following steps: 1. Create multiple cross-validation folds from the data. 2. Filter each fold using the filtering criteria applied to the entire dataset. 3. Run IntLIM for all folds. 4. Process the results for all folds.

Description

Runs the cross-validation end-to-end using the following steps: 1. Create multiple cross-validation folds from the data. 2. Filter each fold using the filtering criteria applied to the entire dataset. 3. Run IntLIM for all folds. 4. Process the results for all folds.

Usage

RunCrossValidation(
  inputData,
  folds,
  analyteType1perc = 0,
  analyteType2perc = 0,
  analyteMiss = 0,
  cov.cutoff = 0,
  stype = "",
  outcome = c(1),
  covar = c(),
  continuous = FALSE,
  save.covar.pvals = FALSE,
  independent.var.type = c(1),
  remove.duplicates = FALSE,
  pvalcutoff = 0.05,
  interactionCoeffPercentile = 0,
  rsquaredCutoff = 0,
  treecuts = 0,
  suppressWarnings = FALSE
)

Arguments

inputData	IntLimData object (output of ReadData()) with analylte levels and associated meta-data
folds	number of folds to create
analyteType1perc	percentile cutoff (0-1) for filtering analyte type 1 (e.g. remove analytes with mean values < 'analyteType1perc' percentile) (default: 0)
analyteType2perc	percentile cutoff (0-1) for filtering analyte type 2 (default: no filtering of analytes) (default:0)
analyteMiss	missing value percent cutoff (0-1) for filtering analytes (analytes with > 80% missing values will be removed) (default:0)
cov.cutoff	percentile cutoff (0-1) for the covariances of the anaytes (default: 0.30)
stype	column name that represents sample type (by default, it will be used in the interaction term). Only 2 categories are currently supported.
outcome	list of outcomes to run. '1' or '2' must be set as outcome/independent variable (default is '1')
covar	Additional variables from the phenotypic data that be integrated into linear model
continuous	boolean to indicate whether the data is continuous or discrete
save.covar.pvals	boolean to indicate whether or not to save the p-values of all covariates, which can be analyzed later but will also lengthen computation time. The default is FALSE.
independent.var.type	list of independent variable types to run. '1' or '2' must be set as independent variable (default is '1')
remove.duplicates	boolean to indicate whether or not to remove the pair with the highest p-value across two duplicate models (e.g. m1~m2 and m2~m1)
pvalcutoff	cutoff of FDR-adjusted p-value for filtering (default 0.05)
interactionCoeffPercentile	percentile cutoff for interaction coefficient
rsquaredCutoff	cutoff for lowest r-squared value
treecuts	user-selected number of clusters (of pairs) to cut the tree into
suppressWarnings	whether to suppress warnings

Value

List of IntResults object with model results (now includes correlations)

IntLIM documentation built on Aug. 22, 2022, 5:05 p.m.