Design-informed size factor estimation

Usage

disize(
  design_formula,
  counts,
  metadata,
  batch_name,
  obs_name = "obs_id",
  n_feats = min(10000L, ncol(counts)),
  n_subset = 50L,
  n_iters = 5000L,
  rel_tol = 1000,
  init_alpha = 1e-08,
  n_threads = 1L,
  n_tries = 5L,
  verbose = 3L
)

Arguments

design_formula: The formula describing the experimental design.
counts: A (observation x feature) count matrix.
metadata: A dataframe containing observation-level metadata.
batch_name: The identifier for the batch column in 'metadata'.
obs_name: The identifier for the observation column in 'metadata'.
n_feats: The number of features used during estimation.
n_subset: The number of observations per experimental unit used during estimation, defaults to 50 (useful for scRNA-seq experiments).
n_iters: The number of iterations used for estimation.
rel_tol: The relative tolerance used for convergence.
init_alpha: The initial step-size for the optimizer, lower values can sometimes make it easier to estimate size factors for more complex designs.
n_threads: The number of threads to use for parallel processing.
n_tries: The maximum number of times to try fitting.
verbose: The verbosity level (1: only errors, 2: also allows warnings,3: also allows messages, 4: also prints additional output useful for debugging).

Value

A named numeric vector containing the size factor estimates.