Multivariate Super Learner Stacking — multivariate

Combines predictions from multiple base methods using optimal weights learned via cross-validation. This meta-learner approach typically improves upon the best single method.

Usage

multivariate_stacking(
  X,
  Y,
  Y.rep = NULL,
  R = NULL,
  id = NULL,
  Omega = NULL,
  base_methods = c("multi_enet", "univariate_enet", "ridge", "lasso"),
  nfolds_stack = 5,
  verbose = FALSE,
  seed = 123,
  parallel = FALSE,
  n.cores = 1
)

Arguments

X: matrix, design matrix of SNP dosages
Y: matrix, matrix of G isoform expression across columns
Y.rep: matrix, matrix of G isoform expression with replicates (optional)
R: int, number of replicates (optional)
id: vector, vector of sample ids (optional)
Omega: matrix, precision matrix of Y (optional, computed if not provided)
base_methods: character vector, methods to use as base learners. Available: "multi_enet", "univariate_enet", "ridge", "lasso", "spls"
nfolds_stack: int, number of CV folds for stacking weights
verbose: logical
seed: int, random seed
parallel: logical, use parallel processing for base methods
n.cores: int, number of cores for parallel processing

Value

isotwas_model object with stacked predictions