mvrt: An R package to generate multivariate t-distributed random data.

Introduction

This package provides random matrices with a defined covariance structure. The package began as an exercise to show performance improvements of using vectorized R code rather than explicit for loops. Later, it became a vehicle for working with the Rcpp and RcppArmadillo packages and learning how to incorporate them into packages of my own. Future work may include FORTRAN code to further explore the capabilities of using compiled code in R scripts and packages.

You can see an exploration the thought process just described in the package vignette. Different methods for function definitions are used and performance is assessed.

Example Usage

library(mvrt)

S <- convert_R2S(make_cor_mat(.9),2:3) # Create a covariance matrix
x <- mvrt( 30, 4:5, S, 29)             # Generate random data with n = 30
y <- mvrt2(30, 4:5, S, 29, .01)        # Random data with maximum abs deviation
                                       # from input parameters specified

# Correlation matrix of x
cor(x)
##           [,1]      [,2]
## [1,] 1.0000000 0.8884765
## [2,] 0.8884765 1.0000000

# Correlation matrix of y
cor(y)
##          [,1]     [,2]
## [1,] 1.000000 0.905187
## [2,] 0.905187 1.000000

Installation

I recommend that you have devtools installed in order to download and install this package. To do so, type the following into your R console:

if(!require(devtools)) install.packages("devtools")
devtools::install_github("pegeler/mvrt")