treediff {treediff}

Perform the treediff test

Description

Perform the treediff test to compare two sets of trees.

Usage

treediff(trees1, trees2, replicates, scale = FALSE, order_labels = FALSE)

## S3 method for class 'treeTest'
print(x, ...)

## S3 method for class 'treeTest'
summary(object, ...)

Arguments

trees1

A list of trees corresponding to the first condition (set). Trees are structured into groups (or clusters) with the same number of replicates in each group. Trees are ordered by groups and then by replicates: {group1+rep1, group1+rep2, ...}. One test is performed for each group.

trees2

A list of trees corresponding to the second condition. Trees are also structured in groups (or clusters) that are exactly the same than for the first condition. The number of replicates in each group can be different from that of trees1.

replicates

A numeric vector of length 2 with the number of replicates for each condition.

scale

Logical. If TRUE, the trees are all rescaled to have a minimum height equal to 0 and a maximum height equal to 1. Default to FALSE.

order_labels

Logical. If TRUE, align leaves ordering in all trees (required if your trees don't have their leaves ordered identically). Default to FALSE.

x

a treeTest object to print

...

not used

object

a treeTest object to print

Details

This function compares two sets of trees using a p-value aggregation method. The p-values are obtained by the treediff method, as described in (Neuvial et al., 2023).

Value

An object of class treeTest with the following entries:

p.value

the p-value for the treediff test.

statistic

the value of the Student's statistic of each leaf pair of the tree test.

p.value.indiv

the p-value of the Student's test for each leaf pair of the tree test.

method

a character string indicating what type of test was performed.

data.name

a character string giving the names of the tree conditions.

Author(s)

Gwendaëlle Cardenas
Marie Chavent [e-mail]
Sylvain Foissac [e-mail]
Pierre Neuvial [e-mail]
Nathanaël Randriamihamison
Nathalie Vialaneix [e-mail]

References

Neuvial Pierre, Randriamihamison Nathanaël, Chavent Marie, Foissac Sylvain and Vialaneix Nathalie (2024) A two-sample tree-based test for hierarchically organized genomic signals. Journal of the Royal Statistical Society, Series C, Forthcoming.

Examples

leaves <- c(100, 120, 50, 80)

trees <- sapply(leaves, FUN = function(leaf) {
  base_data <- matrix(rnorm(2000), nrow = leaf, ncol = 200)

## generates two sets of trees with 4 clusters with 100, 120, 50 and 80
## leaves respectively
## 4 replicates in the first condition and 6 in the second condition

  set1 <- replicate(4, sample(1:100, 50, replace = FALSE))
  set2 <- replicate(6, sample(101:200, 50, replace = FALSE))

  trees1 <- apply(set1, 2, function(asample) {
    samples <- base_data[, asample]
    out <- hclust(dist(samples), method = "ward.D2")
    return(out)
  })

  trees2 <- apply(set2, 2, function(asample) {
    samples <- base_data[, asample]
    out <- hclust(dist(samples), method = "ward.D2")
    return(out)
  })
  return(list("trees1" = trees1, "trees2" = trees2))
})

trees1 <- unlist(trees[1, ], recursive = FALSE)
trees2 <- unlist(trees[2, ], recursive = FALSE)
replicates <- c(4, 6)

tree_pvals <- treediff(trees1, trees2, replicates)
## 4 p-values, one for each cluster
tree_pvals$p.value