Actions	Code Coverage	Website	Doxygen	CRAN Downloads	CRAN

R-Universe	CRAN Packages	DOI

lhs

lhs provides a number of methods for creating and augmenting Latin Hypercube Samples and Orthogonal Array Latin Hypercube Samples.

• Reverse Dependency Checks
  • Depends
  • Imports
  • Suggests
• Docker Images for Testing
  • lhs-debug
  • lhs-revdep built from here

Installation

You can install the released version of lhs from CRAN with:

install.packages("lhs")

You can also install the development version of lhs from github with:

if (!require(devtools)) install.packages("devtools")
devtools::install_github("bertcarnell/lhs")

Quick Start

Create a random LHS with 10 samples and 3 variables:

require(lhs)

## Loading required package: lhs

set.seed(1776)
X <- randomLHS(n = 10, k = 3)

Create a design that is more optimal than the random case:

X_gen <- geneticLHS(10, 3, pop = 100, gen = 5, pMut = 0.1)
X_max1 <- maximinLHS(10, 3, method = "build", dup = 5)
X_max2 <- maximinLHS(10, 3, method = "iterative", optimize.on = "result", eps = 0.01, maxIter = 300)
X_imp <- improvedLHS(10, 3, dup = 5)
X_opt <- optimumLHS(10, 3, maxSweeps = 10, eps = 0.01)

	Method	Mean Distance	Minimum Distance
6	optimum	0.7289	0.4598
2	genetic	0.7190	0.4059
4	maximin	0.7246	0.3975
5	improved	0.7028	0.3872
3	maximin	0.7296	0.3611
1	random	0.7067	0.2709

Augment an existing design:

Y <- randomLHS(10, 5)
Z <- augmentLHS(Y, 2)
dim(Z)

## [1] 12  5

Build an orthogonal array LHS:

# a 9 row design is returned because a 10 row design is not possible with these algorithms
W9 <- create_oalhs(10, 3, bChooseLargerDesign = FALSE, bverbose = FALSE)
dim(W9)

## [1] 9 3

# a 16 row design is returned because a 10 row design is not possible with these algorithms
W16 <- create_oalhs(10, 3, bChooseLargerDesign = TRUE, bverbose = FALSE)
dim(W16)

## [1] 16  3

Create a sliced design where each slice is a Latin hypercube and the union is a Latin hypercube:

S <- slicedLHS(m = 4, t = 3, k = 2)
dim(S)

## [1] 12  2

attr(S, "slices")

##  [1] 1 1 1 1 2 2 2 2 3 3 3 3

Create a nested design where the small design is a subset of the large design:

N <- nestedLHS(small = 4, large = 12, k = 2)
dim(N$large)

## [1] 12  2

all(N$small == N$large[1:4, ])

## [1] TRUE

Help

R-Help Examples of using the LHS package

• Latin hyper cube sampling from expand.grid()
• Latin Hypercube Sampling with a condition
• Latin Hypercube with condition sum = 1
• Latin hypercube sampling
• Latin Hypercube Sample and transformation to uniformly distributed integers or classes
• Latin hypercube sampling from a non-uniform distribution
• Latin Hypercube Sampling when parameters are defined according to specific probability distributions

StackExchange Examples:

• Latin Hypercube around set points
• Latin hypercube sampling with categorical variables
• Are Latin hypercube samples uncorrelated
• Stopping rule for Latin hypercube sampling (LHS)
• Is a group of random hypercube samples equivalent to a single latin hypercube with more samples?
• Taking samples of data using Latin Hypercube Sampling
• Number of parameter sets generated by latin hyercube sampling
• Is there a way to check if sample obeys the Latin Hypercube Sampling rule?
• Effectiveness of Latin Hypercube Sampling
• Dividing CDF rather than PDF equally in Latin Hypercube Sampling
• Stratified sampling / QMC simulation for compound Poisson rv
• Using Latin Hypercube Sampling with a condition that the sum of two variables should be less than one
• How to generate a design for a response surface with a discrete input random variable?
• Is it necessary to shuffle X coordinates in Latin hypercube Sampling?

Other

lhs package announcement: R-pkgs New R-Packages: Triangle and LHS