The package GNLHD. This package provides constructions for LHD, NLHD and GNLHD in R. The package mainly includes construction for GNLHD, which is a kind of new design originated from LHD, and two types of algrorithms to optimizing GNLHDs. Currently, the R package can be only run on Windows platform. More details will be found in a coming paper, Generalized Nested Latin Hypercube Design And Its Optimization written by Daijun Chen and Shifeng Xiong. ##Installation You can also install the development version from Github, which provides daily build of GNLHD:
>install.packages("devtools")
>library(devtools)
>install.packages("numbers")
>library(numbers)
>install_github("DavidDJChen/GNLHD")If you know GIT and R CMD build, it's handly way:
git clone http://github.com/DavidDJChen/GNLHD.git
R CMD BUILD GNLHD
R CMD INSTALL GNLHD_*.tar.gz#Usage Part 1: Construction for GNLHD and some useful functions
>library(GNLHD)
# Construct a GNLHD class u1 with stucture s=(3,5) and q=2.
>u1<-GNLHD$new(s=c(3,5),q=2)
# Show the GNLHD class u1.
>u1
<GNLHD>
Public:
diagnose: function
GNLH: function
GNLH_Full: function
GNLH_illegal_set: function
GNLH_permutation: function
initialize: function
Lcm: 15
q: 2
s: 3 5
StandGNLHD: function
Swap: function
t: 5 3
# Generate a generalzied nested permutation with structure s=(3,5).
> u1$GNLH_permutation()
[1] 10 14 2 9 4 7 15 6 1 13 3 5 8 11 12
# Generate q generalized nested permutations with structure s=(3,5).
>u1$GNLH_Full
[,1] [,2]
[1,] 1 4
[2,] 14 9
[3,] 8 13
[4,] 10 1
[5,] 6 11
[6,] 15 15
[7,] 2 12
[8,] 7 10
[9,] 5 8
[10,] 12 3
[11,] 9 6
[12,] 13 2
[13,] 11 5
[14,] 4 7
[15,] 3 14
# Generate a GNLH structure s=(3,5) and q=2.
> u1$GNLH()
[,1] [,2]
[1,] 4 4
[2,] 13 8
[3,] 9 15
[4,] 10 1
[5,] 3 10
# Generate a Standard GNLHD with s=(3,5) and q=2.
>u1$StandGNLHD()
[,1] [,2]
[1,] 0.2333333 0.4333333
[2,] 0.5666667 0.3000000
[3,] 0.7666667 0.7666667
[4,] 0.9000000 0.1000000
[5,] 0.1000000 0.8333333
# Do layer-in swap operation on a in the 2nd layer and the 1st column.
> a<-u1$GNLH_Full()
> a
[,1] [,2]
[1,] 12 12
[2,] 2 3
[3,] 9 9
[4,] 14 6
[5,] 5 14
[6,] 1 1
[7,] 10 7
[8,] 15 11
[9,] 7 4
[10,] 6 15
[11,] 8 5
[12,] 4 8
[13,] 3 13
[14,] 13 2
[15,] 11 10
> u1$Swap(a, structure=c(3,5),column=1, Swap_type="in", Swap_layer=2,lcm=15)
[,1] [,2]
[1,] 12 12
[2,] 2 3
[3,] 9 9
[4,] 5 6
[5,] 14 14
[6,] 1 1
[7,] 10 7
[8,] 15 11
[9,] 7 4
[10,] 6 15
[11,] 8 5
[12,] 4 8
[13,] 3 13
[14,] 13 2
[15,] 11 10
# Do layer-between swap operation on a in the 2nd layer and the 1st column.
> u1$Swap(a, structure=c(3,5),column=1, Swap_type="between", Swap_layer=2,lcm=15)
[,1] [,2]
[1,] 12 12
[2,] 2 3
[3,] 9 9
[4,] 14 6
[5,] 6 14
[6,] 1 1
[7,] 10 7
[8,] 15 11
[9,] 7 4
[10,] 5 15
[11,] 8 5
[12,] 4 8
[13,] 3 13
[14,] 13 2
[15,] 11 10Part 2: Calucate the Phi_p value with p=50 and t=2 for a.
> Phi_p(Design=a,t=2,p=50)
[1] 0.7169776Part 3: Utilizing an efficient Sequential algorithm to get a optimal GNLHD
# Utilizing an efficient sequential algorithm with iteration=1000 to get a optimal GNLHD
from a.
> Optimal_GNLHD_SequentialAlg(GNLHD=u1, GNLH_Full=a, iteration=1000,T_h_initial=0.1,M=100,J=6,t=2,p=50,
tolerance=0.1,alpha=c(0.8,0.9,0.7))
[,1] [,2]
[1,] 12 12
[2,] 2 3
[3,] 9 9
[4,] 15 4
[5,] 4 15
[6,] 1 1
[7,] 10 7
[8,] 14 11
[9,] 7 5
[10,] 6 14
[11,] 8 6
[12,] 5 8
[13,] 3 13
[14,] 13 2
[15,] 11 10
This package is free and open source software, licensed under GPL-3.
Daijun Chen and Shifeng Xiong, Generalized Nested Latin Hypercube Design and Its Optimization.*1. Deploy on CRAN:
You can install the advanced version on CRAN:
install.packages('GNLHD')*2. Deploy on RForge:
You can also install from RForge,which provides daily build of GNLHD;
# update all existing packages
update.packages(ask=FALSE, repos='http://cran.rstudio.org')
install.packages('GNLHD', repos=c('http://rforge.net','http://cran.rstudio.org'),type='source')