Compute test statistics for matrices

Usage

# Default S3 method
compute_test_statistic(
  input_obj,
  posterior_var_mat,
  case_individuals,
  control_individuals,
  individual_vec,
  min_cells_per_individual = 3,
  verbose = 0,
  ...
)

Arguments

input_obj: Posterior mean matrix (a matrix) where the \(n\) rows represent cells and \(p\) columns represent genes. The rows and columns of the matrix should be named.
posterior_var_mat: Posterior variance matrix (a matrix) where the \(n\) rows represent cells and \(p\) columns represent genes. The rows and columns of the matrix should be the same as those in input_obj.
case_individuals: Vector of strings: the values of individual_vec that are the case individuals.
control_individuals: Vector of strings: the values of individual_vec that are the control individuals.
individual_vec: Vector of strings of length \(n\) (i.e., the number of cells) that denote which cell originates from which individual.
min_cells_per_individual: Minimum number of cells an individual must contribute for the test to be computed. Individuals with fewer are expected to have been dropped upstream by eSVD_helper; reaching here with one is an error rather than a silently noisy statistic. Set to 0 to disable the check.
verbose: Integer.
...: Additional parameters.

Value

A list of named numeric vectors, each of length ncol(input_obj) and carrying colnames(input_obj): teststat_vec, case_mean, control_mean, case_var, control_var, log2fc_se_vec and log2fc_vec. The two variances are those of the mixture over individuals that summarizes each arm, \(\mathrm{mean}_i(\bar V_i) + \mathrm{var}_i(m_i)\), where \(m_i\) and \(\bar V_i\) are individual \(i\)'s average posterior mean and variance and the variance over individuals divides by their number, with no Bessel correction. The statistic is (case_mean - control_mean) / sqrt(case_var / n1 + control_var / n0) for n1 case and n0 control individuals. See compute_log_fold_change for the last two.