R/genmvnorm.R

In gendata: Generate and Modify Synthetic Datasets

genmvnorm <- function(cor, k, n, seed = FALSE){
  
  
  if(is.matrix(cor) == FALSE){
    x <- length(cor)
    if(x != (k * (k - 1) / 2)){stop("STOP: wrong correlation table")}
    
    cr.cor <- matrix(NA, k, k)
    diag(cr.cor) <- 1
    cr.cor[lower.tri(cr.cor)] <- cor
    cr.cor[upper.tri(cr.cor)] <- t(cr.cor)[upper.tri(cr.cor)]
    
    e <- eigen(cr.cor)
    L <- e$values #placing the eigenvalues in V
    Vm <- matrix(0, nrow = k, ncol = k) #creating a k x k matrix.
    diag(Vm) <- L #putting the eigenvalues on the diagonals
    Vm #check-- matrix with eigenvalues
    e$vectors #these are the eigenvectors
    l <- e$vectors %*% sqrt(Vm) #these are the loadings
  }
  
  if(is.matrix(cor) == TRUE){ #if a correlation matrix was used
    
    e <- eigen(cor)
    L <- e$values #placing the eigenvalues in V
    Vm <- matrix(0, nrow = k, ncol = k) #creating a k x k matrix.
    diag(Vm) <- L #putting the eigenvalues on the diagonals
    Vm #check-- matrix with eigenvalues
    e$vectors #these are the eigenvectors
    l <- e$vectors %*% sqrt(Vm) #these are the loadings
    
  }
  
  if(seed != FALSE){set.seed(seed)} 
  f <- matrix(nrow = k, l)
  #dim(f)
  ma <- matrix(nrow = n, ncol = k)
  for (i in 1:k){
    ma[,i] <- rnorm(n)
  }
  tma <- t(ma)
  sol <- f %*% tma
  sol <- t(sol)
  
  data <- data.frame(sol)
  return(data)
  
}