ccdf_testing: Main function to perform complex hypothesis testing using...
In ccdf: Distribution-Free Single-Cell Differential Expression Analysis

Description Usage Arguments Value References Examples

View source: R/ccdf_testing.R

Main function to perform complex hypothesis testing using (un)conditional independence test

ccdf_testing(
  exprmat = NULL,
  variable2test = NULL,
  covariate = NULL,
  distance = c("L2", "L1", "L_sup"),
  test = c("asymptotic", "permutations", "dist_permutations"),
  method = c("linear regression", "logistic regression", "RF"),
  fast = TRUE,
  n_perm = 100,
  n_perm_adaptive = c(100, 150, 250, 500),
  thresholds = c(0.1, 0.05, 0.01),
  parallel = TRUE,
  n_cpus = NULL,
  adaptive = FALSE,
  space_y = FALSE,
  number_y = ncol(exprmat)
)

`exprmat`	a data frame of size `G x n` containing the preprocessed expressions from `n` samples (or cells) for `G` genes. Default is `NULL`.
`variable2test`	a data frame of numeric or factor vector(s) of size `n` containing the variable(s) to be tested (the condition(s))
`covariate`	a data frame of numeric or factor vector(s) of size `n` containing the covariate(s)
`distance`	a character string indicating which distance to use to compute the test, either `'L2'`, `'L1'` or `'L_sup'`, when `method` is `'dist_permutations'`, Default is `'L2'`.
`test`	a character string indicating which method to use to compute the test, either `'asymptotic'`, `'permutations'` or `'dist_permutations'`. `'dist_permutations'` allows to compute the distance between the CDF and the CCDF or two CCDFs. Default is `'asymptotic'`.
`method`	a character string indicating which method to use to compute the CCDF, either `'linear regression'`, `'logistic regression'` and `'permutations'` or `'RF'` for Random Forests. Default is `'linear regression'` since it is the method used in the test.
`fast`	a logical flag indicating whether the fast implementation of logistic regression should be used. Only if `'dist_permutations'` is specified. Default is `TRUE`.
`n_perm`	the number of permutations. Default is `100`.
`n_perm_adaptive`	a vector of the increasing numbers of adaptive permutations when `adaptive` is `TRUE`. `length(n_perm_adaptive)` should be equal to `length(thresholds)+1`. Default is `c(0.1,0.05,0.01)`.
`thresholds`	a vector of the decreasing thresholds to compute adaptive permutations when `adaptive` is `TRUE`. `length(thresholds)` should be equal to `length(n_perm_adaptive)-1`. Default is `c(100,150,250,500)`.
`parallel`	a logical flag indicating whether parallel computation should be enabled. Default is `TRUE`.
`n_cpus`	an integer indicating the number of cores to be used when `parallel` is `TRUE`. Default is `parallel::detectCores() - 1`.
`adaptive`	a logical flag indicating whether adaptive permutations should be performed. Default is `FALSE`.
`space_y`	a logical flag indicating whether the y thresholds are spaced. When `space_y` is `TRUE`, a regular sequence between the minimum and the maximum of the observations is used. Default is `FALSE`.
`number_y`	an integer value indicating the number of y thresholds (and therefore the number of regressions) to perform the test. Default is `ncol(exprmat)`.

A list with the following elements:

which_test: a character string carrying forward the value of the 'which_test' argument indicating which test was performed (either 'asymptotic','permutations','dist_permutations').
n_perm: an integer carrying forward the value of the 'n_perm' argument or 'n_perm_adaptive' indicating the number of permutations performed (NA if asymptotic test was performed).
pval: computed p-values. A data frame with one raw for each gene, and with 2 columns: the first one 'raw_pval' contains the raw p-values, the second one 'adj_pval' contains the FDR adjusted p-values using Benjamini-Hochberg correction.

Gauthier M, Agniel D, Thiébaut R & Hejblum BP (2019). Distribution-free complex hypothesis testing for single-cell RNA-seq differential expression analysis, *bioRxiv* 445165. [DOI: 10.1101/2021.05.21.445165](https://doi.org/10.1101/2021.05.21.445165).

X <- as.factor(rbinom(n=100, size = 1, prob = 0.5))
Y <- t(replicate(10, ((X==1)*rnorm(n = 50,0,1)) + ((X==0)*rnorm(n = 50,0.5,1))))
res_asymp <- ccdf_testing(exprmat=data.frame(Y=Y), 
variable2test=data.frame(X=X), test="asymptotic",
n_cpus=1)$pvals # asymptotic test