R/datasim.R

In gaow/mmbr: Multivariate Sum of Single Effects Regression

#' @title Simulate mvSuSiE Data
#'
#' @description A simple simulation function to simulate some test
#'   data.
#'
#' @param n Number of samples to simulate.
#'
#' @param p Number of features to simulate.
#'
#' @param r Number of conditions to simulate.
#'
#' @param s Number of effect variables per condition if greater than
#'   1; otherwise, proportion of effect variables per condition if less
#'   than 1.
#'
#' @param center_scale Describe what input argument "center" is for.
#'
#' @param y_missing Describe what input argument "y_missing" is for.
#'
#' @return Describe the outputs here.
#'
#' @importFrom stats cov
#' @importFrom stats runif
#' @importFrom stats rnorm
#'
#' @export
#'
mvsusie_sim1 <- function(n = 200, p = 500, r = 2, s = 4,
                         center_scale = FALSE, y_missing = NULL) {
  X <- matrix(rnorm(n * p, 0, 1), n, p)
  if (s >= 1) {
    beta <- matrix(0, p, r)
    for (i in 1:r) {
      beta[sample(1:p, s), i] <- 1
    }
  } else {
    beta <- matrix(runif(p * r) > s, p, r)
  }
  y <- X %*% beta + do.call(cbind, lapply(1:r, function(i) rnorm(n)))
  if (center_scale) {
    X <- scale(X)
    y <- t(t(y) - apply(y, 2, mean))
  }
  if (!is.null(y_missing)) {
    y2 <- y
    for (i in 1:nrow(y2)) {
      y2[i, runif(r) <= y_missing] <- NA
    }
    y_missing <- y2
  }
  scaled_prior_variance <- 0.2
  return(list(
    X = X, y = y, y_missing = y_missing, d = diag(t(X) %*% X),
    n = n, p = p, r = r, V = scaled_prior_variance * cov(y),
    b = beta
  ))
}