simmr: Similarity-based multimodal regression

In andy1764/SiMMR: Similarity-Based Multimodal Regression

Similarity-based multimodal regression

Description

simmr fits a multimodal regression model using only distance matrices computed on the original data. Permutation testing is used to obtain p-values.

Usage

simmr(
  D,
  X,
  variables,
  tests = c("Dempster", "PC"),
  D.scale = c("max", "trace", "dvar", "none"),
  pc.pv = NULL,
  pc.n = NULL,
  n.perm = 999,
  eigen.tol = 1e-10,
  debug = FALSE
)

Arguments

D	List of distance matrices of class matrix or dist
X	Design matrix, either output of model.matrix or valid input to the function.
variables	Names of explanatory variables to regress on, must match with column names of X
tests	Chosen test statistics from "Dempster", "PC". For "PC", defaults to using n-p-1 PCs, where X is ⁠n x p⁠. If pc.pv and pc.n are both specified, pc.pv is prioritized.
D.scale	Method for normalization of distance matrices
pc.pv	Optional, calculate additional PC-based test statistics for specified proportions of variation explained
pc.n	Optional, calculate additional PC-based test statistics for specific numbers of PCs. If 'all', tries every possible number of PCs.
n.perm	Number of permutations to perform
eigen.tol	Discard eigenvalues below this tolerance for both PCA and computing test statistics
debug	Include internal objects in output (from unpermuted analysis)

Value

A list with the following components:

stat	Test statistics computed on original data.
perms	Test statistics after permutations.
perm.p	Permutation p-values for each test.

Examples

D <- list(dist(rnorm(10)), dist(rnorm(10)), dist(rnorm(10)))
X <- list("var" = runif(10))
simmr(D, X, "var")

andy1764/SiMMR documentation built on Jan. 29, 2024, 1:28 a.m.