R 体素明智的Mann-Whitney U/独立样本T检验

我试图用R，以体素的方式，计算和显示Mann-Whitney U测试中的p值，这类似于更熟悉的独立样本t测试，在.nii文件数据上。我从Joset a.Etzel在2014年写的博客中找到了一个很好的演示代码，可以用于一个样本t测试。该代码的摘录如下所示：

library(oro.nifti) 
library(plyr)

rm(list=ls()) 

in.path <- "VoxelThings/"   # where the input NIfTI images are
out.path <- "VoxelThings/demo_outputs/"  # where to write the NIfTI t-value image
sub.ids <- c("a", "b", "c", "d")   # subject ID codes, corresponding to the input images
img.dims <- c(91,109,91)  # expected image dimensions. it would be more elegant coding to read the size out of the input images.

getT <- function(x) {    # function to calculate the t-test at each voxel and return the t value
  # we can't calculate a t-test if variance is zero, so check before trying.
  if (var(x) == 0) { stat <- NA; } 
  
  else { stat <- t.test(x, alternative="greater", mu=0.5)$p.value; }
  
  return(stat)
}

# load the images for each person into a big data structure
big <- array(NA, c(img.dims, length(sub.ids)));  # 3d array for each person, 4th dimension are the subjects
for (sid in 1:length(sub.ids)) {   # sid <- 1;
  img <- readNIfTI(paste0(in.path, sub.ids[sid], "_demo.nii.gz"));
  if (dim(img)[1] != img.dims[1] | dim(img)[2] != img.dims[2] | dim(img)[3] != img.dims[3]) { stop("not the expected dimensions"); }
  big[,,,sid] <- img;
}


big.out <- aaply(big, c(1,2,3), getT);   # this line calls the t-test function at each voxel. aaply isn't particularly fast, but is easy to code.
if (dim(big.out)[1] != img.dims[1] | dim(big.out)[2] != img.dims[2] | dim(big.out)[3] != img.dims[3]) { stop("not the expected big.out dims"); }

或者在我真正想要的Mann-Whitney U测试中，它是这样的：

wilcox.test(dependent_variable ~ grouping_variable, data = df, exact = TRUE, correct = TRUE)

这两种测试都需要将数据指定到一个组中；例如，a&b可以与c&d相比较——我在这里怎么做？ 非常感谢您的帮助，谢谢

wilcox.test(dependent_variable ~ grouping_variable, data = df, exact = TRUE, correct = TRUE)