In mpadge/tnorm: Truncated Normal Distribution

tnorm

tnorm is an R package for generating random samples from truncated normal distributions. It exists solely as a faster replacement for the core functionality of other packages, and it useful in simulation contexts where speed is important and samples from truncated normal distributions are required. This functionality already exists in both

1. msm::rtnorm, which is entirely R-based, and

2. truncnorm::rtruncnorm, which has core code in C.

Even the latter of these is, however, slower than tnorm at generating random values. tnorm currently generates distributions centred around a mean value of 1, truncated to between 0 and 2. Values can be readily re-scaled to desired mean values and limits.

---

Install

devtools::install_github ('mpadge/tnorm')

while (length (grep ('tnorm', getwd ())) > 0) setwd ("..")
#devtools::document ("tnorm")
#devtools::check ("tnorm")
#Rcpp::compileAttributes ("tnorm")
devtools::load_all ("tnorm")
setwd ("./tnorm")
#testthat::test_package ('tnorm')

---

Use

The one and only function

z <- tnormn (n=1e5, sd=1)

Compare timing with msm::rtnorm and truncnorm::rtruncnorm

library (rbenchmark)
n <- 1e6
mn <- sd <- 1
knitr::kable (benchmark (
           tnorm::tnormn (n = n, sd = sd),
           msm::rtnorm (n = n, mean = mn, sd = sd, lower = 0, upper = 2),
           truncnorm::rtruncnorm (n = n, sd = sd, a = 0, b = 2),
           replications = 10,
           order = "relative") [, 1:4])

mpadge/tnorm documentation built on Nov. 22, 2019, 1:34 a.m.