Hopefully taking out the complexity of using the simplex.
The simplexity package contains functions to generate, manipulate and plot data on the simplex
(platform independent) If you have devtools installed you can use:
library(devtools) # see https://www.r-project.org/nosvn/pandoc/devtools.html
devtools::install_github('tystan/simplexity') # see DESCRIPTION file for required and suggested packages(it will compile some C++ code contained in the package for your platform)
If you don't have devtools installed but are using Windows, you can download and install the below zip file:
simplexity_0.2.2.zip (C++ code is pre-compiled, no compiler required)
library(simplexity)
# now package loaded, see help file to run example
?enumerate_simplex
# or create an example 4-simplex plot
example("plot_four_comp", package = "simplexity")
# or look at the ellipsoid fencing vignette
vignette("ellipsoid-fencing", package = "simplexity")# create a grid of evenly spaced simplex values
# remove observations that are on the edge of the simplex (see examples of `enumerate_simplex()`)
D <- 4; n_chunks <- 5;
grid_4simplex <- enumerate_simplex(D, n_chunks - D) + 1
# There will be 4 rows and 4 cols in the returned matrix
# This matrix will be approximately 0 Mb in memory
grid_4simplex <- grid_4simplex / n_chunks # apply closure row sums = 1
colnames(grid_4simplex) <- paste0("comp", 1:ncol(grid_4simplex))
grid_4simplex
# comp1 comp2 comp3 comp4
# [1,] 0.4 0.2 0.2 0.2
# [2,] 0.2 0.4 0.2 0.2
# [3,] 0.2 0.2 0.4 0.2
# [4,] 0.2 0.2 0.2 0.4
# simulate a fictitious response variable based on 4-simplex ilrs
y <- simplex_to_ilr(grid_4simplex) %*% matrix(c(-1, 1, 0.5), ncol = 1)
colnames(y) <- "outcome"
# outcome
# [1,] -0.6002831
# [2,] 0.7660467
# [3,] 0.1621827
# [4,] -0.3279463
# add the response variable to the simplex data for plotting
grid_4simplex <- as.data.frame(cbind(y, grid_4simplex))
# this is an interactive/movable plotly 3D scatterplot
plot_four_comp(grid_4simplex, "comp1", "comp2", "comp3", "comp4", col = "outcome")
