The goal of DoEAna is to visualize the results of research regarding space-filling experimental designs. The package has a prepared data set “DoEAna.rda” with designs for comparison. In the dataset, you can see how designs were generated on two different geometries: hypercube and simplex. They also differ by the distance measure used: Manhattan, Chebyshev, and Euclidean. Finally, they differ by the number of design points: 10, 20, and 30. The user can use this package to compare how these different experiment features affect a designs performance. Although the main purpose of this package is to display specific results of a research project, these functions can be used by any individual that formats their data in the same way as DoEAna.rda in the package data folder.
You can install the development version of DoEAna from GitHub with:
# install.packages("devtools")
devtools::install_github("DianaF65/DoEAna")This is a basic example which shows you how to use this package:
First, select optimal space-filling designs. You can view the design matrix and plot them.
library(DoEAna)
# select designs on the hypercube or simplex
my_design_hyper <- select_design(d_data = sf_designs, geometry = "hypercube", distance = "chebyshev", n = 20)
my_design_sim <- select_design(d_data = sf_designs, geometry = "simplex", distance = "chebyshev", n = 20)
# show the design matrices
print_design(my_design_hyper)
#> [,1] [,2]
#> [1,] -0.241240 -0.497962
#> [2,] 0.997182 0.001212
#> [3,] 0.195633 0.500962
#> [4,] -0.914856 -0.497256
#> [5,] -0.436894 -0.997358
#> [6,] 0.971241 -0.498390
#> [7,] -0.999211 -0.999125
#> [8,] 0.951683 -0.997436
#> [9,] -0.223096 0.001856
#> [10,] 0.461232 0.001846
#> [11,] 0.334787 -0.498980
#> [12,] -0.425508 0.999994
#> [13,] 0.323609 0.999992
#> [14,] -0.457074 0.501036
#> [15,] 0.304456 -0.999984
#> [16,] -0.999801 0.501026
#> [17,] -0.995570 0.002004
#> [18,] 0.887859 0.500900
#> [19,] -0.999597 0.999997
#> [20,] 0.888506 0.999958
print_design(my_design_sim)
#> [,1] [,2] [,3]
#> [1,] 0.009484 0.194663 0.795853
#> [2,] 0.582070 0.224648 0.193282
#> [3,] 0.245236 0.603631 0.151133
#> [4,] 0.600517 0.029450 0.370033
#> [5,] 0.795879 0.002175 0.201946
#> [6,] 0.000000 1.000000 0.000000
#> [7,] 0.992595 0.001118 0.006287
#> [8,] 0.182242 0.801162 0.016596
#> [9,] 0.214295 0.000764 0.784941
#> [10,] 0.430589 0.000007 0.569404
#> [11,] 0.209907 0.376617 0.413475
#> [12,] 0.384869 0.402233 0.212899
#> [13,] 0.006165 0.396049 0.597786
#> [14,] 0.453662 0.538201 0.008138
#> [15,] 0.230711 0.181308 0.587981
#> [16,] 0.019140 0.767096 0.213764
#> [17,] 0.800675 0.199325 0.000000
#> [18,] 0.081351 0.571904 0.346745
#> [19,] 0.405351 0.204038 0.390611
#> [20,] 0.000000 0.000449 0.999551
# plot the designs
par(mfrow = c(1, 2))
plot_design(my_design_hyper)plot_design(my_design_sim)
#> Registered S3 methods overwritten by 'ggtern':
#> method from
#> grid.draw.ggplot ggplot2
#> plot.ggplot ggplot2
#> print.ggplot ggplot2Investigate their relative efficiency.
# compare across different distance measures with relative efficiency on the
# hypercube
my_designM_H <- select_design(d_data = sf_designs, geometry = "hypercube",
distance = "manhattan", n = 20)
my_designE_H <- select_design(d_data = sf_designs, geometry = "hypercube",
distance = "euclidean", n = 20)
my_designC_H <- select_design(d_data = sf_designs, geometry = "hypercube",
distance = "chebyshev", n = 20)
relative_efficiency(m_row = my_designM_H, e_row = my_designE_H,
c_row = my_designC_H)
#> M Best E Best C Best
#> M(x) 100.000 77.529 71.639
#> E(x) 91.230 100.000 92.378
#> C(x) 69.923 76.814 100.000# compare across different distance measures with relative efficiency on the
# simplex
my_designM_S <- select_design(d_data = sf_designs, geometry = "simplex",
distance = "manhattan", n = 20)
my_designE_S <- select_design(d_data = sf_designs, geometry = "simplex",
distance = "euclidean", n = 20)
my_designC_S <- select_design(d_data = sf_designs, geometry = "simplex",
distance = "chebyshev", n = 20)
relative_efficiency(m_row = my_designM_S, e_row = my_designE_S,
c_row = my_designC_S)
#> M Best E Best C Best
#> M(x) 100.000 98.665 99.376
#> E(x) 88.071 100.000 88.735
#> C(x) 100.628 99.285 100.000Or view an FDS plot
# Compare relative prediction variance across designs using FDS plots
fds_plot(design1 = my_designC_H, design2 = my_designM_H,
title = "Chebychev vs Manahattan Distance for n=20")