In CollinErickson/sFFLHD: Sequential Full Factorial-Based Latin Hypercube Design

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "tools/README-"
)

sFFLHD

This R package provides a class that generates experiment sFFLHD designs. Sequential full factorial-based Latin hypercube design were created by Duan, Ankenman, Sanchez, and Sanchez (2015, Technometrics).

To create a new design you use the function sFFLHD$new and must give in the number of dimensions, D, and the batch size/number of levels per factor, L. An example is shown below (the last line can be repeated when run in console to see how new batches are added).

library(sFFLHD)
set.seed(0)
s <- sFFLHD$new(D=2,L=3)
plot(s$get.batch(),xlim=0:1,ylim=0:1,pch=19)
abline(h=(0:(s$Lb))/s$Lb,v=(0:(s$Lb))/s$Lb,col=3);points(s$get.batch(),pch=19)

By default the new points are selected using maximin distance optimization to spread them out. This is why points will end up near corners. This option will slow down the code a little but generally not noticeably compared to what the design is used for. If set to FALSE then the points are randomly placed within their small grid box.

CollinErickson/sFFLHD documentation built on May 6, 2019, 12:28 p.m.