Perform False Discovery Rate (FDR) Control

This function performs hypothesis testing by controlling the False Discovery Rate (FDR) based on the local false discovery rate (lfdr) stored in the fash object.

fdr_control(fash_obj, alpha = 0.05, plot = FALSE, sort = FALSE)

Arguments

fash_obj: A fash object containing the results of the FASH pipeline, including the vector lfdr.
alpha: A numeric value specifying the FDR threshold (significance level) used to declare discoveries. Default is 0.05.
plot: A logical value. If TRUE, plots the cumulative FDR values against the rank of units sorted by lfdr, with a horizontal line at the alpha level. Default is FALSE.
sort: A logical value. If TRUE, returns the FDR results sorted by increasing lfdr.

Value

A list with the following components:

fdr_results

A data frame with one row per unit, containing:

rank: Rank of the unit after sorting by lfdr (1 = smallest lfdr).
index: Original index of the unit in lfdr.
lfdr: The sorted local false discovery rate values.
FDR: The cumulative FDR at each rank, computed as the running mean of the sorted lfdr values.

message

A character string summarizing how many units are significant at the chosen alpha level and the total number of units tested.

significant_units

A vector of indices (or names, if lfdr is named) corresponding to units with cumulative FDR less than or equal to alpha.

Examples

# Example usage
set.seed(1)
data_list <- list(
  data.frame(y = rnorm(n = 5, sd = 0.5), x = 1:5, offset = 0),
  data.frame(y = rnorm(n = 5, sd = 0.8), x = 1:5, offset = 0),
  data.frame(y = rnorm(n = 5, sd = 0.6), x = 1:5, offset = 0),
  data.frame(y = rnorm(n = 5, sd = 0.7), x = 1:5, offset = 0)
)
S <- list(rep(0.5, 5), rep(0.8, 5), rep(0.6, 5), rep(0.7, 5))
grid <- seq(0, 2, length.out = 10)
fash_obj <- fash(data_list = data_list, Y = "y", smooth_var = "x", offset = "offset", S = S, grid = grid, likelihood = "gaussian", verbose = TRUE)
#> Starting data setup...
#> Completed data setup in 0.00 seconds.
#> Starting likelihood computation...
#> 
  |                                                                            
  |                                                                      |   0%
  |                                                                            
  |==================                                                    |  25%
  |                                                                            
  |===================================                                   |  50%
  |                                                                            
  |====================================================                  |  75%
  |                                                                            
  |======================================================================| 100%
#> Completed likelihood computation in 0.10 seconds.
#> Starting empirical Bayes estimation...
#> Completed empirical Bayes estimation in 0.00 seconds.
#> fash object created successfully.
fdr_control(fash_obj, alpha = 0.05, plot = TRUE)
#> 0 datasets are significant at alpha level 0.05. Total datasets tested: 4. 

#> $fdr_results
#>           rank index lfdr FDR
#> Dataset_1    1     1    1   1
#> Dataset_2    2     2    1   1
#> Dataset_3    3     3    1   1
#> Dataset_4    4     4    1   1
#> 
#> $message
#> [1] "0 datasets are significant at alpha level 0.05. Total datasets tested: 4."
#> 
#> $significant_units
#> character(0)
#>