Simulate multi-omics data with known braid patterns — simulate_braid

Generates matched sample-level data for method development and validation. Residuals are correlated across omic layers for each entity, allowing the covariance, heterogeneity, equivalence, and classification procedures to be tested against known cross-layer effects.

Usage

simulate_braid_data(
  n_per_group = 60L,
  n_reference = NULL,
  n_comparison = NULL,
  omics = c("RNA", "Protein", "Metabolite"),
  patterns = NULL,
  rho = 0.4,
  missing_rate = 0,
  modality_missing_rate = 0,
  residual_distribution = c("normal", "t"),
  t_df = 5,
  seed = 1L
)

Arguments

n_per_group: Default samples per group when `n_reference` and `n_comparison` are not supplied.
n_reference: Optional reference-group sample size.
n_comparison: Optional comparison-group sample size.
omics: Ordered omic names.
patterns: Named list of true standardized mean shifts, one numeric vector per entity.
rho: Scalar equicorrelation or an omic-by-omic residual correlation matrix.
missing_rate: Independent value-level missingness probability.
modality_missing_rate: Probability of an entire subject modality being absent; scalar or named by omic.
residual_distribution: `"normal"` or heavy-tailed `"t"` residuals.
t_df: Degrees of freedom for t residuals; must exceed 2 for finite variance.
seed: Random seed.

Value

List with `data` (`omics_braid_data`) and `truth` table.