generateMed: Generate a high-dimensional mediation dataset

In vankesteren/cmfilter: Coordinate-Wise Mediation Filter

Generate a high-dimensional mediation dataset

Description

This function generates a dataset from an x -> M -> y model, where M may be of any size with any correlation matrix.

Usage

generateMed(
  n = 100L,
  a = 0.3,
  b = 0.3,
  r2y = 0.5,
  dir = 0,
  Sigma,
  residual = FALSE,
  empirical = FALSE,
  scaley = FALSE,
  forma = identity,
  formb = identity
)

Arguments

n	Sample size
a	Vector of a path coefficients within 0 and 1
b	Vector of b path coefficients within 0 and 1
r2y	Proportion of explained variance in y. Set to b %*% Sigma %*% b for var(y) == 1.
dir	Direct path from x to y
Sigma	Desired true covariance matrix between the mediators M
residual	Whether Sigma indicates residual or marginal covariance
empirical	Ensure observed data matrix has exactly the requested covmat (only if Sigma is specified)
scaley	Whether to standardise y (changes b path coefficients)
forma	Functional form of the a paths. Function that accepts a matrix as input and transforms each column to the desired form.
formb	Functional form of the b paths. Function that accepts a vector.

Value

A data frame with columns x, M.1 - M.p, y

Examples

# Generate a suppression dataset where M.2 is suppressed
sup <- generateMed(n = 100,
                   a = c(-0.4, 0.4),
                   b = c(0.8, 0.48),
                   Sigma = matrix(c(1, -0.6, -0.6, 1), 2))

vankesteren/cmfilter documentation built on April 6, 2023, 3:40 a.m.