R/mn_cmvbtld.R

In tharris0924/btld: Triangular and Bimodal Triangular Linked Distributions

#' Random generation from the compound Multinomial-MVBTLD.
#'
#' This function is a random generator for length n (d)-dimensional MVBTLD random variables. It also returns the density of these values.
#'
#'
#' @param n size of dataset desired
#' @param size integer indicating the total number of objects to draw from in the multinomial
#' @param alpha scale matrix. Must be d-by-2 dimensions.
#' @param theta mode matrix. Must be d-by-2 dimensions.
#' @param sigma value on the off diagonal for Gaussian covariance matrix used in copula generation.
#' @param dim number of dimensions of copula. Must be same dimensions as the mode and scale parameters.
#' @param base the base value to use in the softmax function
#' @param ... kws for \code{normalCopula} and \code{rCopula}.
#' @return matrix n by d+1 random variates from the compound MN-MVBTLD.
#'
#' @importFrom stats rmultinom dmultinom
#' @export
#' @examples
#' alpha<-matrix(c(3,3,3,3),nrow=2,ncol=2, byrow=TRUE)
#' theta<-matrix(c(0.3,0.7,0.3,0.7),ncol=2,nrow=2,byrow=TRUE)
#' mn_cmvbtld(100, size=10, alpha=alpha, theta=theta, sigma=0.5,dim=2,base=1000)

mn_cmvbtld<-function(n,size,alpha,theta,sigma,dim,base,...){
  df <- matrix(nrow=n,ncol=dim)
  for(i in seq_len(n)){
    p <- rmvbtld(1, alpha = alpha,theta = theta,sigma=sigma,dim = dim,...)
    p<-softmax(p,exp=F,base=base)
    df[i,] <- t(rmultinom(n=1,size = size,prob=p))
  }
  return(df)
}