Zigzag expanded navigation plots in R: The R package zenplots

M. Hofert and R. W. Oldford

2019-08-01

This vignette accompanies the paper “Zigzag expanded navigation plots in R: The R package zenplots”. Note that sections are numbered accordingly (or omitted). Furthermore, it is recommended to read the paper to follow this vignette.

# attaching required packages
library(PairViz)
library(MASS)
## Warning: package 'MASS' was built under R version 3.5.2
library(zenplots)

2 Zenplots

As example data, we use the olive data set:

data(olive, package = "zenplots")

Reproducing the plots of Figure 1:

zenplot(olive)

zenplot(olive, plot1d = "layout", plot2d = "layout")

Considering the str()ucture of zenplot() (here formatted for nicer output):

str(zenplot)
function (x, turns = NULL, first1d = TRUE, last1d = TRUE,
          n2dcols = c("letter", "square", "A4", "golden", "legal"),
          n2dplots = NULL,
          plot1d = c("label", "points", "jitter", "density", "boxplot",
                     "hist", "rug", "arrow", "rect", "lines", "layout"),
          plot2d = c("points", "density", "axes", "label", "arrow",
                     "rect", "layout"),
          zargs = c(x = TRUE, turns = TRUE, orientations = TRUE,
                    vars = TRUE, num = TRUE, lim = TRUE, labs = TRUE,
                    width1d = TRUE, width2d = TRUE,
                    ispace = match.arg(pkg) != "graphics"),
          lim = c("individual", "groupwise", "global"),
          labs = list(group = "G", var = "V", sep = ", ", group2d = FALSE),
          pkg = c("graphics", "grid", "loon"),
          method = c("tidy", "double.zigzag", "single.zigzag"),
          width1d = if (is.null(plot1d)) 0.5 else 1,
          width2d = 10,
          ospace = if (pkg == "loon") 0 else 0.02,
          ispace = if (pkg == "graphics") 0 else 0.037, draw = TRUE, ...)

2.1 Layout

To investigate the layout options of zenplots a bit more, we need a larger data set. To this end we simply double the olive data here (obviously only for illustration purposes):

olive2 <- cbind(olive, olive) # just for this illustration

Reproducing the plots of Figure 2:

zenplot(olive2, n2dcols = 6, plot1d = "layout", plot2d = "layout",
        method = "single.zigzag")

zenplot(olive2, n2dcols = 6, plot1d = "layout", plot2d = "layout",
        method = "double.zigzag")

zenplot(olive2, n2dcols = 6, plot1d = "layout", plot2d = "layout",
        method = "tidy")

Note that there is also method = "rectangular" (leaving the zigzagging zenplot paradigm but being useful for laying out 2d plots which are not necessarily connected through a variable; note that in this case, we omit the 1d plots as the default (labels) is rather confusing in this example):

zenplot(olive2, n2dcols = 6, plot1d = "arrow", plot2d = "layout",
        method = "rectangular")

Reproducing the plots of Figure 3:

zenplot(olive, plot1d = "layout", plot2d = "layout", method = "double.zigzag",
        last1d = FALSE, ispace = 0.1)

zenplot(olive, plot1d = "layout", plot2d = "layout", n2dcol = 4, n2dplots = 8,
        width1d = 2, width2d = 4)

3 Zenpaths

A very basic path (standing for the sequence of pairs (1,2), (2,3), (3,4), (4,5)):

(path <- 1:5)
## [1] 1 2 3 4 5

A zenpath through all pairs of variables (Eulerian):

(path <- zenpath(5))
##  [1] 5 1 2 3 1 4 2 5 3 4 5

If dataMat is a five-column matrix, the zenplot of all pairs would then be constructed as follows:

zenplot(x = dataMat[,path])

The str()ucture of zenpath() (again formatted for nicer output):

str(zenpath)
function (x, pairs = NULL,
          method = c("front.loaded", "back.loaded", "balanced",
                     "eulerian.cross", "greedy.weighted", "strictly.weighted"),
          decreasing = TRUE)

Here are some methods for five variables:

zenpath(5, method = "front.loaded")
##  [1] 5 1 2 3 1 4 2 5 3 4 5
zenpath(5, method = "back.loaded")
##  [1] 1 2 3 1 4 2 5 3 4 5 1
zenpath(5, method = "balanced")
##  [1] 1 2 3 5 4 1 3 4 2 5 1

The following method considers two groups: One of size three, the other of size five. The sequence of pairs is constructed such that the first variable comes from the first group, the second from the second.

zenpath(c(3,5), method = "eulerian.cross")
##  [1] 1 4 2 5 1 6 2 7 1 8 3 4 3 6 7 3 5 8 2

Reproducing Figure 4:

oliveAcids <- olive[, !names(olive) %in% c("area", "region")] # acids only
zpath <- zenpath(ncol(oliveAcids)) # all pairs
zenplot(oliveAcids[, zpath], plot1d = "hist", plot2d = "density")

4 Build your own zenplots

4.3 Custom layout and plots – a spiral of ggplots example

Figure 5 can be reproduced as follows (note that we do not show the plot here due to a CRAN issue when running this vignette):

path <- c(1,2,3,1,4,2,5,1,6,2,7,1,8,2,3,4,5,3,6,4,7,3,8,4,5,6,7,5,8,6,7,8)
turns <- c("l",
           "d","d","r","r","d","d","r","r","u","u","r","r","u","u","r","r",
           "u","u","l","l","u","u","l","l","u","u","l","l","d","d","l","l",
           "u","u","l","l","d","d","l","l","d","d","l","l","d","d","r","r",
           "d","d","r","r","d","d","r","r","d","d","r","r","d","d")

library(ggplot2) # for ggplot2-based 2d plots
stopifnot(packageVersion("ggplot2") >= "2.2.1") # need 2.2.1 or higher
ggplot2d <- function(zargs) {
  r <- extract_2d(zargs)
  num2d <- zargs$num/2
  df <- data.frame(x = unlist(r$x), y = unlist(r$y))
  p <- ggplot() +
    geom_point(data = df, aes(x = x, y = y), cex = 0.1) +
    theme(axis.line = element_blank(),
          axis.ticks = element_blank(),
          axis.text.x = element_blank(),
          axis.text.y = element_blank(),
          axis.title.x = element_blank(),
          axis.title.y = element_blank())
  if(num2d == 1) p <- p +
    theme(panel.background = element_rect(fill = 'royalblue3'))
  if(num2d == (length(zargs$turns)-1)/2) p <- p +
    theme(panel.background = element_rect(fill = 'maroon3'))
  ggplot_gtable(ggplot_build(p))
}

zenplot(as.matrix(oliveAcids)[,path], turns = turns, pkg = "grid",
        plot2d = function(zargs) ggplot2d(zargs))

4.4 Data groups

Split the olive data set into three groups (according to their variable area):

oliveAcids.by.area <- split(oliveAcids, f = olive$area)
# Replace the "." by " " in third group's name
names(oliveAcids.by.area)[3] <- gsub("\\.", " ", names(oliveAcids.by.area)[3])
names(oliveAcids.by.area)
## [1] "South"        "Sardinia"     "Centre North"

Reproducing the plots of Figure 6 (note that lim = "groupwise" does not make much sense here as a plot):

zenplot(oliveAcids.by.area, labs = list(group = NULL))

zenplot(oliveAcids.by.area, lim = "groupwise", labs = list(sep = " - "),
        plot1d = function(zargs) label_1d_graphics(zargs, cex = 0.8),
        plot2d = function(zargs)
            points_2d_graphics(zargs, group... = list(sep = "\n - \n")))

4.5 Custom zenpaths

Find the “convexity” scagnostic for each pair of olive acids.

library(scagnostics)
Y <- scagnostics(oliveAcids) # compute scagnostics (scatter-plot diagonstics)
X <- Y["Convex",] # pick out component 'convex'
d <- ncol(oliveAcids)
M <- matrix(, nrow = d, ncol = d) # matrix with all 'convex' scagnostics
M[upper.tri(M)] <- X # (i,j)th entry = scagnostic of column pair (i,j) of oliveAcids
M[lower.tri(M)] <- t(M)[lower.tri(M)] # symmetrize
round(M, 5)
##         [,1]    [,2]    [,3]    [,4]    [,5]    [,6]    [,7]    [,8]
## [1,]      NA 0.48952 0.46343 0.45887 0.43914 0.34583 0.31259 0.28413
## [2,] 0.48952      NA 0.42276 0.50499 0.44591 0.35855 0.35846 0.31729
## [3,] 0.46343 0.42276      NA 0.39700 0.36394 0.31316 0.29534 0.33709
## [4,] 0.45887 0.50499 0.39700      NA 0.46454 0.36616 0.29451 0.34888
## [5,] 0.43914 0.44591 0.36394 0.46454      NA 0.31977 0.31443 0.36750
## [6,] 0.34583 0.35855 0.31316 0.36616 0.31977      NA 0.53726 0.34001
## [7,] 0.31259 0.35846 0.29534 0.29451 0.31443 0.53726      NA 0.22231
## [8,] 0.28413 0.31729 0.33709 0.34888 0.36750 0.34001 0.22231      NA

Show the six pairs with largest “convexity” scagnostic:

zpath <- zenpath(M, method = "strictly.weighted") # list of ordered pairs
head(M[do.call(rbind, zpath)]) # show the largest six 'convexity' measures
## [1] 0.5372599 0.5049945 0.4895179 0.4645377 0.4634277 0.4588675

Extract the corresponding pairs:

(ezpath <- extract_pairs(zpath, n = c(6, 0))) # extract the first six pairs
## [[1]]
## [1] 7 6
## 
## [[2]]
## [1] 4 2
## 
## [[3]]
## [1] 2 1
## 
## [[4]]
## [1] 5 4
## 
## [[5]]
## [1] 3 1
## 
## [[6]]
## [1] 4 1

Reproducing Figure 7 (visualizing the pairs):

library(graph)
plot(graph_pairs(ezpath)) # depict the six most convex pairs (edge = pair)

Connect them:

(cezpath <- connect_pairs(ezpath)) # keep the same order but connect the pairs
## [[1]]
## [1] 7 6
## 
## [[2]]
## [1] 4 2 1
## 
## [[3]]
## [1] 5 4
## 
## [[4]]
## [1] 3 1 4

Build the corresponding list of matrices:

oliveAcids.grouped <- groupData(oliveAcids, indices = cezpath) # group data for (zen)plotting

Reproducing Figure 8 (zenplot of the six pairs of acids with largest “convexity” scagnostic):

zenplot(oliveAcids.grouped)