ggmulti: High Dimensional Data Visualization

Description

It provides materials (i.e. 'serial axes' objects, Andrew's plot, various glyphs for scatter plot) to visualize high dimensional data.

Author(s)

Maintainer: Zehao Xu z267xu@uwaterloo.ca

Authors:

• R. Wayne Oldford rwoldford@uwaterloo.ca

Base Geom ggproto classes for ggplot2

Description

All geom_ functions (like geom_point) return a layer that contains a Geom object (like GeomPoint). The Geom object is responsible for rendering the data in the plot. Each of the Geom objects is a ggproto object, descended from the top-level Geom, and each implements various methods and fields. Compared to Stat and Position, Geom is a little different because the execution of the setup and compute functions is split up. setup_data runs before position adjustments, and draw_layer is not run until render time, much later. This means there is no setup_params because it's hard to communicate the changes.

Usage

GeomDensity_

GeomBar_

GeomImageGlyph

GeomPolygonGlyph

GeomQuantiles

GeomSerialaxesDensity

GeomSerialAxesGlyph

GeomSerialaxesHist

GeomSerialaxesQuantile

GeomSerialaxes

Format

An object of class GeomDensity_ (inherits from GeomRibbon, Geom, ggproto, gg) of length 6.

An object of class GeomBar_ (inherits from GeomBar, GeomRect, Geom, ggproto, gg) of length 4.

An object of class GeomImageGlyph (inherits from Geom, ggproto, gg) of length 7.

An object of class GeomPolygonGlyph (inherits from Geom, ggproto, gg) of length 7.

An object of class GeomQuantiles (inherits from GeomQuantile, GeomPath, Geom, ggproto, gg) of length 1.

An object of class GeomSerialaxesDensity (inherits from GeomDensity_, GeomRibbon, Geom, ggproto, gg) of length 2.

An object of class GeomSerialAxesGlyph (inherits from Geom, ggproto, gg) of length 7.

An object of class GeomSerialaxesHist (inherits from GeomBar_, GeomBar, GeomRect, Geom, ggproto, gg) of length 2.

An object of class GeomSerialaxesQuantile (inherits from GeomPath, Geom, ggproto, gg) of length 4.

An object of class GeomSerialaxes (inherits from GeomPath, Geom, ggproto, gg) of length 3.

NBA 30 Teams Statistics in 20-21 Regular Season

Description

A dataset containing the statistics (e.g. Points Per Game, Average Field Goals Made, etc) of 30 NBA Teams in 2020-2021 regular season

Format

A data frame with 30 rows (teams) and 42 variables:

Team

Team Names.

CONF

Factor; Conference of Teams (West or East).

DIV

Factor; Division of Teams.

Playoff

Factor; Whether Teams are in (0 or 1) Playoffs.

PTS

Points Per Game.

FGM

Average Field Goals Made.

FGA

Average Field Goals Attempted.

FG%

Field Goal Percentage.

3PM

Average 3-Point Field Goals Made.

3PA

Average 3-Point Field Goals Attempted.

3P%

3-Point Field Goal Percentage.

FTM

Average Free Throws Made.

FTA

Average Free Throws Attempted.

FT%

Free Throw Percentage.

OR

Offensive Rebounds Per Game.

DR

Defensive Rebounds Per Game.

REB

Rebounds Per Game.

AST

Assists Per Game.

STL

Steals Per Game.

BLK

Blocks Per Game.

TO

Turnovers Per Game.

PF

Fouls Per Game.

OPTS

Opponent Points Per Game.

OFGM

Opponent Average Field Goals Made.

OFGA

Opponent Average Field Goals Attempted.

OFG%

Opponent Field Goal Percentage.

O3PM

Opponent Average 3-Point Field Goals Made.

O3PA

Opponent Average 3-Point Field Goals Attempted.

O3P%

Opponent 3-Point Field Goal Percentage.

OFTM

Opponent Average Free Throws Made.

OFTA

Opponent Average Free Throws Attempted.

OFT%

Opponent Free Throw Percentage.

OOR

Opponent Offensive Rebounds Per Game.

ODR

Opponent Defensive Rebounds Per Game.

OREB

Opponent Rebounds Per Game.

OAST

Opponent Assists Per Game.

OSTL

Opponent Steals Per Game.

OBLK

Opponent Blocks Per Game.

OTO

Opponent Turnovers Per Game.

OPF

Opponent Fouls Per Game.

Win

Win Games in Regular Season.

Lose

Loss Games in Regular Season.

Author(s)

Zehao Xu

Source

https://www.espn.com/nba/stats/team/_/season/2021

Base Position ggproto classes for ggplot2

Description

All position_ functions (like position_dodge) return a Position object (like PositionDodge). The Position object is responsible for adjusting the position of overlapping geoms. The way that the position_ functions work is slightly different from the geom_ and stat_ functions, because a position_ function actually "instantiates" the Position object by creating a descendant, and returns that. Each of the Position objects is a ggproto object, descended from the top-level Position.

Usage

PositionDodge_

PositionDodge2_

PositionIdentity_

PositionStack_

PositionFill_

Format

An object of class PositionDodge_ (inherits from PositionDodge, Position, ggproto, gg) of length 2.

An object of class PositionDodge2_ (inherits from PositionDodge2, PositionDodge, Position, ggproto, gg) of length 2.

An object of class PositionIdentity_ (inherits from PositionIdentity, Position, ggproto, gg) of length 3.

An object of class PositionStack_ (inherits from PositionStack, Position, ggproto, gg) of length 3.

An object of class PositionFill_ (inherits from PositionStack_, PositionStack, Position, ggproto, gg) of length 2.

Base Stat ggproto classes for ggplot2

Description

All stat_ functions (like stat_bin()) return a layer that contains a Stat object (like StatBin). The Stat object is responsible for rendering the data in the plot. Each of the Stat objects is a ggproto object, descended from the top-level Stat, and each implements various methods and fields.

Usage

StatDensity_

StatHist_

StatBin_

StatCount_

StatSerialaxesDensity

StatSerialaxesHist

StatSerialaxes

StatDotProduct

Format

An object of class StatDensity_ (inherits from StatDensity, Stat, ggproto, gg) of length 4.

An object of class StatHist_ (inherits from StatBin, Stat, ggproto, gg) of length 4.

An object of class StatBin_ (inherits from StatHist_, StatBin, Stat, ggproto, gg) of length 2.

An object of class StatCount_ (inherits from StatHist_, StatBin, Stat, ggproto, gg) of length 2.

An object of class StatSerialaxesDensity (inherits from StatDensity, Stat, ggproto, gg) of length 4.

An object of class StatSerialaxesHist (inherits from StatBin, Stat, ggproto, gg) of length 4.

An object of class StatSerialaxes (inherits from Stat, ggproto, gg) of length 6.

An object of class StatDotProduct (inherits from StatSerialaxes, Stat, ggproto, gg) of length 4.

Layers for serial axes coordinate

Description

Project the regular geom layers onto the serial axes coordinate.

Usage

add_serialaxes_layers(layer, plot, object, axes)

Arguments

Details

The class is determined by layers you add. For example, you want to add a boxplot layer on serial axes coordinate. By the ggplot syntax, it should be ggplot(data, mapping) + geom_boxplot() + coord_serialaxes() To make it work, object add_serialaxes_layers.GeomBoxplot must be created. In this function, some computations will be applied.

coord_radar (deprecated)

Description

use coord_radial

Usage

coord_radar(theta = "x", start = 0, direction = 1, clip = "on")

Arguments

Radial axes

Description

A radial (spider) coordinate. A wrapper of the function coord_polar() by forcing it linear.

Usage

coord_radial(theta = "x", start = 0, direction = 1, clip = "on")

Arguments

Details

The serial histogram and serial density cannot be applied on a radial coordinate yet.

Examples

if(require("dplyr")) {
ggplot(NBAstats2021, mapping = aes(colour = Playoff)) +
  geom_serialaxes(
    axes.sequence = c("PTS", "OPTS", "3PM", "O3PM", "PTS"),
      scaling = "variable"
    ) +
  coord_radial() +
  scale_x_continuous(
    breaks = 1:5,
    labels = c("Points",
               "Oppo Points",
               "3P Made",
               "Oppo 3P Made",
               "Points Per Game")) +
  scale_y_continuous(labels = NULL) +
  facet_wrap(~CONF)
  }

Serial axes coordinates

Description

It is mainly used to visualize the high dimensional data set either on the parallel coordinate or the radial coordinate.

Usage

coord_serialaxes(
  axes.layout = c("parallel", "radial"),
  scaling = c("data", "variable", "observation", "none"),
  axes.sequence = character(0L),
  positive = TRUE,
  ...
)

Arguments

Details

Serial axes coordinate system (parallel or radial) is different from the Cartesian coordinate system or its transformed system (say polar in ggplot2) since it does not have a formal transformation (i.e. in polar coordinate system, "x = rcos(theta)", "y = rsin(theta)"). In serial axes coordinate system, mapping aesthetics does not really require "x" or "y". Any "non-aesthetics" components passed in the mapping system will be treated as an individual axis.

To project a common geom layer on such serialaxes, users can customize function add_serialaxes_layers.

Value

a ggproto object

Potential Risk

In package ggmulti, the function ggplot_build.gg is provided. At the ggplot construction time, the system will call ggplot_build.gg first. If the plot input is not a CoordSerialaxes coordinate system, the next method ggplot_build.ggplot will be called to build a "gg" plot; else some geometric transformations will be applied first, then the next method ggplot_build.ggplot will be executed. So, the potential risk is, if some other packages e.g. foo, also provide a function ggplot_build.gg that is used for their specifications but the namespace is beyond the ggmulti (ggmulti:::ggplot_build.gg is covered), error may occur. If so, please consider using the geom_serialaxes.

Examples

if(require("dplyr")) {
# Data
nba <- NBAstats2021 %>%
  mutate(
    dPTS = PTS - OPTS,
    dREB = REB - OREB,
    dAST = AST - OAST,
    dTO = TO - OTO
  )
# set sequence by `axes.sequence`
p <- ggplot(nba,
            mapping = aes(
              dPTS = dPTS,
              dREB = dREB,
              dAST = dAST,
              dTO = dTO,
              colour = Win
            )) +
       geom_path(alpha = 0.2) +
       coord_serialaxes(axes.layout = "radial") +
       scale_color_gradient(low="blue", high="red")
p
# quantile layer
p + geom_quantiles(quantiles = c(0.5),
                   colour = "green", linewidth = 1.2)

# facet
p +
  facet_grid(Playoff ~ CONF)
}

Transformation Coefficients

Description

The dimension of the original data set is n\*p. It can be projected onto a n\*k space. The functions below are to provide such transformations, e.g. the Andrews coefficient (a Fourier transformation) and the Legendre polynomials.

Usage

andrews(p = 4, k = 50 * (p - 1), ...)

legendre(p = 4, k = 50 * (p - 1), ...)

Arguments

Value

A list contains two named components

1. vector: A length k vector (define the domain)

2. matrix: A p\*k transformed coefficient matrix

References

Andrews, David F. "Plots of high-dimensional data." Biometrics (1972): 125-136.

Abramowitz, Milton, and Irene A. Stegun, eds. "Chapter 8" Handbook of mathematical functions with formulas, graphs, and mathematical tables. Vol. 55. US Government printing office, 1948.

Examples

x <- andrews(p = 4)
dat <- iris[, -5]
proj <- t(as.matrix(dat) %*% x$matrix)
matplot(x$vector, proj,
        type = "l", lty = 1,
        col = "black",
        xlab = "x",
        ylab = "Andrews coefficients",
        main = "Iris")

More general smoothed density estimates

Description

Computes and draws kernel density estimate. Compared with geom_density(), it provides more general cases that accepting x and y. See details

Usage

geom_density_(
  mapping = NULL,
  data = NULL,
  stat = "density_",
  position = "identity_",
  ...,
  scale.x = NULL,
  scale.y = c("data", "group", "variable"),
  as.mix = FALSE,
  positive = TRUE,
  prop = 0.9,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

stat_density_(
  mapping = NULL,
  data = NULL,
  geom = "density_",
  position = "stack_",
  ...,
  bw = "nrd0",
  adjust = 1,
  kernel = "gaussian",
  n = 512,
  trim = FALSE,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

Arguments

Details

The x (or y) is a group variable (categorical) and y (or x) is the target variable (numerical) to be plotted. If only one of x or y is provided, it will treated as a target variable and ggplot2::geom_density will be executed.

There are four combinations of scale.y and as.mix.

scale.y = "group" and as.mix = FALSE

The density estimate area of each subgroup (represented by each color) within the same group is the same.

scale.y = "group" and as.mix = TRUE

The density estimate area of each subgroup (represented by each color) within the same group is proportional to its own counts.

scale.y = "data" and as.mix = FALSE

The sum of density estimate area of all groups is scaled to maximum of 1. and the density area for each group is proportional to the its count. Within each group, the area of each subgroup is the same.

scale.y = "data" and as.mix = TRUE

The sum of density estimate area of all groups is scaled to maximum of 1 and the area of each subgroup (represented by each color) is proportional to its own count.

See vignettes[https://great-northern-diver.github.io/ggmulti/articles/histogram-density-.html] for more intuitive explanation.

Orientation

This geom treats each axis differently and, thus, can thus have two orientations. Often the orientation is easy to deduce from a combination of the given mappings and the types of positional scales in use. Thus, ggplot2 will by default try to guess which orientation the layer should have. Under rare circumstances, the orientation is ambiguous and guessing may fail. In that case the orientation can be specified directly using the orientation parameter, which can be either "x" or "y". The value gives the axis that the geom should run along, "x" being the default orientation you would expect for the geom.

See Also

geom_density, geom_hist_

Examples

if(require(dplyr)) {
  mpg %>%
    dplyr::filter(drv != "f") %>%
    ggplot(mapping = aes(x = drv, y = cty, fill = factor(cyl))) +
    geom_density_(alpha = 0.1)

  # only `x` or `y` is provided
  # that would be equivalent to call function `geom_density()`
  diamonds %>%
    dplyr::sample_n(500) %>%
    ggplot(mapping = aes(x = price)) +
    geom_density_()

  # density and boxplot
  # set the density estimate on the left
  mpg %>%
    dplyr::filter(drv != "f") %>%
    ggplot(mapping = aes(x = drv, y = cty,
                         fill = factor(cyl))) +
    geom_density_(alpha = 0.1,
                  scale.y = "group",
                  as.mix = FALSE,
                  positive = FALSE) +
    geom_boxplot()

  # x as density
  set.seed(12345)
  suppressWarnings(
    diamonds %>%
      dplyr::sample_n(500) %>%
      ggplot(mapping = aes(x = price, y = cut, fill = color)) +
      geom_density_(orientation = "x", prop = 0.25,
                    position = "stack_",
                    scale.y = "group")
  )
}
# settings of `scale.y` and `as.mix`

ggplots <- lapply(list(
                      list(scale.y = "data", as.mix = TRUE),
                      list(scale.y = "data", as.mix = FALSE),
                      list(scale.y = "group", as.mix = TRUE),
                      list(scale.y = "group", as.mix = FALSE)
                    ),
                   function(vars) {
                     scale.y <- vars[["scale.y"]]
                     as.mix <- vars[["as.mix"]]
                     ggplot(mpg,
                            mapping = aes(x = drv, y = cty, fill = factor(cyl))) +
                       geom_density_(alpha = 0.1, scale.y = scale.y, as.mix = as.mix) +
                       labs(title = paste("scale.y =", scale.y),
                            subtitle = paste("as.mix =", as.mix))
                   })
suppressWarnings(
  gridExtra::grid.arrange(grobs = ggplots)
)

More general histogram

Description

More general histogram (geom_histogram) or bar plot (geom_bar). Both x and y could be accommodated. See details

Usage

geom_hist_(
  mapping = NULL,
  data = NULL,
  stat = "hist_",
  position = "stack_",
  ...,
  scale.x = NULL,
  scale.y = c("data", "group", "variable"),
  as.mix = FALSE,
  binwidth = NULL,
  bins = NULL,
  positive = TRUE,
  prop = 0.9,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

geom_histogram_(
  mapping = NULL,
  data = NULL,
  stat = "bin_",
  position = "stack_",
  ...,
  scale.x = NULL,
  scale.y = c("data", "group"),
  as.mix = FALSE,
  positive = TRUE,
  prop = 0.9,
  binwidth = NULL,
  bins = NULL,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

geom_bar_(
  mapping = NULL,
  data = NULL,
  stat = "count_",
  position = "stack_",
  ...,
  scale.x = NULL,
  scale.y = c("data", "group"),
  positive = TRUE,
  prop = 0.9,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

stat_hist_(
  mapping = NULL,
  data = NULL,
  geom = "bar_",
  position = "stack_",
  ...,
  binwidth = NULL,
  bins = NULL,
  center = NULL,
  boundary = NULL,
  breaks = NULL,
  closed = c("right", "left"),
  pad = FALSE,
  width = NULL,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

stat_bin_(
  mapping = NULL,
  data = NULL,
  geom = "bar_",
  position = "stack_",
  ...,
  binwidth = NULL,
  bins = NULL,
  center = NULL,
  boundary = NULL,
  breaks = NULL,
  closed = c("right", "left"),
  pad = FALSE,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

stat_count_(
  mapping = NULL,
  data = NULL,
  geom = "bar_",
  position = "stack_",
  ...,
  width = NULL,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

Arguments

Details

x (or y) is a group variable (categorical) and y (or x) a target variable (numerical) to be plotted. If only one of x or y is provided, it will treated as a target variable and ggplot2::geom_histogram will be executed. Several things should be noticed:

1. If both x and y are given, they can be one discrete one continuous or two discrete. But they cannot be two continuous variables (which one will be considered as a group variable?).

2. geom_hist_ is a wrapper of geom_histogram_ and geom_count_. Suppose the y is our interest (x is the categorical variable), geom_hist_() can accommodate either continuous or discrete y. While, geom_histogram_() only accommodates the continuous y and geom_bar_() only accommodates the discrete y.

3. There are four combinations of scale.y and as.mix.

scale.y = "group" and as.mix = FALSE

The density estimate area of each subgroup (represented by each color) within the same group is the same.

scale.y = "group" and as.mix = TRUE

The density estimate area of each subgroup (represented by each color) within the same group is proportional to its own counts.

scale.y = "data" and as.mix = FALSE

The sum of density estimate area of all groups is scaled to maximum of 1. and the density area for each group is proportional to the its count. Within each group, the area of each subgroup is the same.

scale.y = "data" and as.mix = TRUE

The sum of density estimate area of all groups is scaled to maximum of 1 and the area of each subgroup (represented by each color) is proportional to its own count.

See vignettes[https://great-northern-diver.github.io/ggmulti/articles/histogram-density-.html] for more intuitive explanation. Note that, if it is a grouped bar chart (both x and y are categorical), parameter 'as.mix' is meaningless.

Orientation

This geom treats each axis differently and, thus, can thus have two orientations. Often the orientation is easy to deduce from a combination of the given mappings and the types of positional scales in use. Thus, ggplot2 will by default try to guess which orientation the layer should have. Under rare circumstances, the orientation is ambiguous and guessing may fail. In that case the orientation can be specified directly using the orientation parameter, which can be either "x" or "y". The value gives the axis that the geom should run along, "x" being the default orientation you would expect for the geom.

See Also

geom_histogram, geom_density_

Examples

if(require(dplyr) && require(tidyr)) {

  # histogram
  p0 <- mpg %>%
    dplyr::filter(manufacturer %in% c("dodge", "ford", "toyota", "volkswagen")) %>%
    ggplot(mapping = aes(x = manufacturer, y = cty))
  p0 + geom_hist_()

  ## set position
  #### default is "stack_"
  p0 + geom_hist_(mapping = aes(fill = fl))
  #### "dodge_"
  p0 + geom_hist_(position = "dodge_",
                  mapping = aes(fill = fl))
  #### "dodge2_"
  p0 + geom_hist_(position = "dodge2_",
                  mapping = aes(fill = fl))

  # bar chart
  mpg %>%
    ggplot(mapping = aes(x = drv, y = class)) +
    geom_hist_(orientation = "y")

  # scale.y as "group"
  p <- iris %>%
    tidyr::pivot_longer(cols = -Species,
                        names_to = "Outer sterile whorls",
                        values_to = "x") %>%
    ggplot(mapping = aes(x = `Outer sterile whorls`,
                         y = x, fill = Species)) +
    stat_hist_(scale.y = "group",
               prop = 0.6,
               alpha = 0.5)
  p
  # with density on the left
  p + stat_density_(scale.y = "group",
                    prop = 0.6,
                    alpha = 0.5,
                    positive = FALSE)

  ########### only `x` or `y` is provided ###########
  # that would be equivalent to call function
  # `geom_histogram()` or `geom_bar()`
  ### histogram
  diamonds %>%
    dplyr::sample_n(500) %>%
    ggplot(mapping = aes(x = price)) +
    geom_hist_()
  ### bar chart
  diamonds %>%
    dplyr::sample_n(500) %>%
    ggplot(mapping = aes(x = cut)) +
    geom_hist_()
}

Add image glyphs on scatter plot

Description

Each point glyph can be an image (png, jpeg, etc) object.

Usage

geom_image_glyph(
  mapping = NULL,
  data = NULL,
  stat = "identity",
  position = "identity",
  ...,
  images,
  imagewidth = 1.2,
  imageheight = 0.9,
  interpolate = TRUE,
  na.rm = FALSE,
  show.legend = NA,
  inherit.aes = TRUE
)

Arguments

Value

a geom layer

Aesthetics

geom_..._glyph() understands the following aesthetics (required aesthetics are in bold):

• x

• y

• alpha

• colour

• fill

• group

• size

• linetype

• shape

• stroke

The size unit is cm

Note that the shape and stroke do not have real meanings unless the essential argument images is missing. If so, a point visual will be displayed with corresponding shape and stroke.

See Also

geom_serialaxes_glyph, geom_polygon_glyph

Examples

# image glyph
if(require("png")) {
img_path <- list.files(file.path(find.package(package = 'ggmulti'),
                                 "images"),
                       full.names = TRUE)
Raptors <- png::readPNG(img_path[2L])
Warriors <- png::readPNG(img_path[3L])

pg <- ggplot(data = data.frame(x = 1:2, y = rep(1, 2)),
       mapping = aes(x = x, y = y)) +
  geom_image_glyph(images = list(Raptors,
                                 Warriors),
                   imagewidth = rep(1.2, 2),
                   imageheight = c(0.9, 1.2)) +
  coord_cartesian(xlim = extendrange(c(1,2)))
pg
# query the images (a numerical array)
build <- ggplot2::ggplot_build(pg)
# `imageRaptors` and `imageWarriors` are three dimensional
# arrays (third dimension specifying the plane)
imageRaptors <- build$data[[1]]$images[[1]]
imageWarriors <- build$data[[1]]$images[[2]]

if(require("grid")) {
grid.newpage()
grid.raster(imageRaptors)
grid.newpage()
grid.raster(imageWarriors)
}

# THIS IS SLOW
mercLogo <- png::readPNG(img_path[1L])

p <- ggplot(mapping = aes(x = hp, y = mpg)) +
       geom_point(
         data = mtcars[!grepl("Merc", rownames(mtcars)), ],
         color = "skyblue") +
       geom_image_glyph(
         data = mtcars[grepl("Merc", rownames(mtcars)), ],
         images = mercLogo,
         imagewidth = 1.5
       )
p

}

Add polygon glyphs on scatter plot

Description

Each point glyph can be a polygon object. We provide some common polygon coords in polygon_glyph. Also, users can customize their own polygons.

Usage

geom_polygon_glyph(
  mapping = NULL,
  data = NULL,
  stat = "identity",
  position = "identity",
  ...,
  polygon_x,
  polygon_y,
  linewidth = 1,
  na.rm = FALSE,
  show.legend = NA,
  inherit.aes = TRUE
)

Arguments

Value

a geom layer

Aesthetics

geom_..._glyph() understands the following aesthetics (required aesthetics are in bold):

• x

• y

• alpha

• colour

• fill

• group

• size

• linetype

• shape

• stroke

The size unit is cm

Note that the shape and stroke do not have real meanings unless the essential argument polygon_x or polygon_y is missing. If so, a point visual will be displayed with corresponding shape and stroke.

See Also

geom_serialaxes_glyph, geom_image_glyph

Examples

# polygon glyph
p <- ggplot(data = data.frame(x = 1:4, y = 1:4),
            mapping = aes(x = x, y = y)) +
  geom_polygon_glyph(polygon_x = list(x_star, x_cross, x_hexagon, x_airplane),
                     polygon_y = list(y_star, y_cross, y_hexagon, y_airplane),
                     colour = 'black', fill = 'red')
p

# the coords of each polygons can be achieved by calling function `ggplot_build`
build <- ggplot2::ggplot_build(p)
polygon_x <- build$data[[1]]$polygon_x
polygon_y <- build$data[[1]]$polygon_y

Add quantile layers on serial axes coordinate

Description

In ggplot2, geom_quantile() is used to fit a quantile regression to the data and draws the fitted quantiles with lines. However, geom_quantiles() is mainly used to draw quantile lines on serial axes. See examples

Usage

geom_quantiles(
  mapping = NULL,
  data = NULL,
  stat = "quantile",
  position = "identity",
  ...,
  lineend = "butt",
  linejoin = "round",
  linemitre = 10,
  na.rm = FALSE,
  show.legend = NA,
  inherit.aes = TRUE
)

Arguments

See Also

geom_serialaxes_quantile

Examples

p <- ggplot(iris,
            mapping = aes(
              Sepal.Length = Sepal.Length,
              Sepal.Width = Sepal.Width,
              Petal.Length = Petal.Length,
              Petal.Width = Petal.Width
            )
  ) +
  geom_path(alpha = 0.2)  +
  coord_serialaxes(scaling = "variable")
p + geom_quantiles(colour = c("red", "green", "blue"),
                   quantiles = c(0.25, 0.5, 0.75),
                   linewidth = 2)

Serial axes layer

Description

Draw a serial axes layer, parallel axes under Cartesian system and radial axes under Polar system. It only takes the "widens" data. Each non-aesthetics component defined in the mapping aes() will be treated as an axis.

Usage

geom_serialaxes(
  mapping = NULL,
  data = NULL,
  stat = "serialaxes",
  position = "identity",
  ...,
  axes.sequence = character(0L),
  merge = TRUE,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

stat_serialaxes(
  mapping = NULL,
  data = NULL,
  geom = "serialaxes",
  position = "identity",
  ...,
  axes.sequence = character(0L),
  merge = TRUE,
  axes.position = NULL,
  scaling = c("data", "variable", "observation", "none"),
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

stat_dotProduct(
  mapping = NULL,
  data = NULL,
  geom = "path",
  position = "identity",
  ...,
  axes.sequence = character(0L),
  merge = TRUE,
  scaling = c("data", "variable", "observation", "none"),
  transform = andrews,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

Arguments

Details

The difference between the "lengthens" data and "widens" data can be found in Tidy Data. How to transform one to the other is explained in tidyr

See Also

coord_serialaxes, geom_serialaxes_density, geom_serialaxes_quantile, geom_serialaxes_hist

Andrews plot andrews, Legendre polynomials legendre

Examples

# parallel coordinate
p <- ggplot(NBAstats2021,
            mapping = aes(FGA = FGA,
                          `3PA` = `3PA`,
                          FTA = FTA,
                          OFGA = OFGA,
                          O3PA = O3PA,
                          OFTA = OFTA,
                          colour = CONF))

# Teams in West are more likely to make 3-point field goals.
# Besides, they have a better performance in restricting opponents
# to make 3-point field goals.
p +
  geom_serialaxes(scaling = "variable",
                  alpha = 0.4,
                  linewidth = 3) +
  scale_x_continuous(breaks = 1:6,
                     labels = c("FGA", "3PA", "FTA",
                                "OFGA", "O3PA", "OFTA")) +
  scale_y_continuous(labels = NULL)

# andrews plot
p + geom_serialaxes(stat = "dotProduct",
                    scaling = "variable",
                    transform = andrews) # default

# Legendre polynomials
p + geom_serialaxes(stat = "dotProduct",
                    scaling = "variable",
                    transform = legendre)


############# Determine axes sequence
# 1. set the duplicated axes by mapping aesthetics
ggplot(iris, mapping = aes(Sepal.Length = Sepal.Length,
                           Sepal.Width = Sepal.Width,
                           Sepal.Length = Sepal.Length,
                           Sepal.Width = Sepal.Width,
                           colour = Species)) +
  # only two axes, duplicated axes are removed
  geom_serialaxes()

# 2. set the duplicated axes by axes.sequence
ggplot(iris, mapping = aes(colour = Species)) +
  geom_serialaxes(
    axes.sequence = c("Sepal.Length", "Sepal.Width",
                      "Sepal.Length", "Sepal.Width"))

Smoothed density estimates for "widens" data under serial axes coordinate

Description

Computes and draws kernel density estimates on serial axes coordinate for each non-aesthetics component defined in the mapping aes().

Usage

geom_serialaxes_density(
  mapping = NULL,
  data = NULL,
  stat = "serialaxes_density",
  position = "identity_",
  ...,
  axes.sequence = character(0L),
  merge = TRUE,
  scale.y = c("data", "group"),
  as.mix = TRUE,
  positive = TRUE,
  prop = 0.9,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

stat_serialaxes_density(
  mapping = NULL,
  data = NULL,
  geom = "serialaxes_density",
  position = "stack_",
  ...,
  axes.sequence = character(0L),
  merge = TRUE,
  axes.position = NULL,
  scaling = c("data", "variable", "observation", "none"),
  bw = "nrd0",
  adjust = 1,
  kernel = "gaussian",
  n = 512,
  trim = FALSE,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

Arguments

See Also

geom_density_, geom_serialaxes, geom_serialaxes_quantile, geom_serialaxes_hist

Examples

p <- ggplot(iris, mapping = aes(Sepal.Length = Sepal.Length,
                                Sepal.Width = Sepal.Width,
                                Petal.Length = Petal.Length,
                                Petal.Width = Petal.Width,
                                colour = Species,
                                fill = Species)) +
       geom_serialaxes(alpha = 0.2) +
       geom_serialaxes_density(alpha = 0.5) +
       scale_x_continuous(breaks = 1:4,
                          labels = colnames(iris)[-5]) +
       scale_y_continuous(labels = NULL) +
       xlab("variable") +
       ylab("") +
       theme(axis.text.x = element_text(angle = 45, vjust = 0.5))
p

Add serial axes glyphs on scatter plot

Description

To visualize high dimensional data on scatterplot. Each point glyph is surrounded by a serial axes (parallel axes or radial axes) object.

Usage

geom_serialaxes_glyph(
  mapping = NULL,
  data = NULL,
  stat = "identity",
  position = "identity",
  ...,
  serialaxes.data,
  axes.sequence = character(0L),
  scaling = c("data", "variable", "observation", "none"),
  axes.layout = c("parallel", "radial"),
  andrews = FALSE,
  show.axes = FALSE,
  show.enclosing = FALSE,
  linewidth = 1,
  axescolour = "black",
  bboxcolour = "black",
  na.rm = FALSE,
  show.legend = NA,
  inherit.aes = TRUE
)

Arguments

Value

a geom layer

Aesthetics

geom_..._glyph() understands the following aesthetics (required aesthetics are in bold):

• x

• y

• alpha

• colour

• fill

• group

• size

• linetype

• shape

• stroke

The size unit is cm

Note that the shape and stroke do not have real meanings unless the essential argument serialaxes.data is missing. If so, a point visual will be displayed with corresponding shape and stroke.

See Also

geom_polygon_glyph, geom_image_glyph

Examples

# serial axes glyph
p <- ggplot(data = iris,
            mapping = aes(x = Sepal.Length, y = Sepal.Width, colour = Species)) +
  geom_serialaxes_glyph(serialaxes.data = iris[, -5],
                        axes.layout = "radial")
p

Histogram for "widens" data under serial axes coordinate

Description

Computes and draws histogram on serial axes coordinate for each non-aesthetics component defined in the mapping aes().

Usage

geom_serialaxes_hist(
  mapping = NULL,
  data = NULL,
  stat = "serialaxes_hist",
  position = "stack_",
  ...,
  axes.sequence = character(0L),
  axes.position = NULL,
  merge = TRUE,
  scale.y = c("data", "group"),
  as.mix = TRUE,
  positive = TRUE,
  prop = 0.9,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

stat_serialaxes_hist(
  mapping = NULL,
  data = NULL,
  geom = "serialaxes_hist",
  position = "stack_",
  ...,
  axes.sequence = character(0L),
  scaling = c("data", "variable", "observation", "none"),
  axes.position = NULL,
  binwidth = NULL,
  bins = NULL,
  center = NULL,
  boundary = NULL,
  breaks = NULL,
  closed = c("right", "left"),
  pad = FALSE,
  width = NULL,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

Arguments

See Also

geom_hist_, geom_serialaxes, geom_serialaxes_quantile, geom_serialaxes_density

Examples

p <- ggplot(NBAstats2021,
            mapping = aes(`FG%` = `FG%`,
                          `3P%` = `3P%`,
                          `FT%` = `FT%`,
                          `OFG%` = `OFG%`,
                          `O3P%` = `O3P%`,
                          `OFT%` = `OFT%`,
                          colour = Playoff,
                          fill = Playoff)) +
            geom_serialaxes(alpha = 0.2,
                            scaling = "variable") +
            geom_serialaxes_hist(alpha = 0.5,
                                 prop = 0.7,
                                 scaling = "variable") +
            scale_x_continuous(breaks = 1:6,
                               labels = c("FG", "3P", "FT",
                                          "OFG", "O3P", "OFT")) +
            scale_y_continuous(labels = NULL) +
            xlab("variable") +
            ylab("") +
            theme(axis.text.x = element_text(angle = 45, vjust = 0.5))
p

Quantile layer for serial axes coordinate

Description

Draw a quantile layer for serial axes coordinate. Don't be confused with geom_quantile() which is a quantile regression. See examples.

Usage

geom_serialaxes_quantile(
  mapping = NULL,
  data = NULL,
  stat = "serialaxes",
  position = "identity",
  ...,
  quantiles = seq(0, 1, 0.25),
  axes.sequence = character(0L),
  merge = TRUE,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

stat_serialaxes_quantile(
  mapping = NULL,
  data = NULL,
  geom = "serialaxes_quantile",
  position = "identity",
  ...,
  axes.sequence = character(0L),
  merge = TRUE,
  quantiles = seq(0, 1, 0.25),
  scaling = c("data", "variable", "observation", "none"),
  axes.position = NULL,
  na.rm = FALSE,
  orientation = NA,
  show.legend = NA,
  inherit.aes = TRUE
)

Arguments

See Also

geom_density_, geom_serialaxes, geom_serialaxes_density, geom_serialaxes_hist

Examples

# lower quantile, median and upper quantile
p <- ggplot(iris, mapping = aes(Sepal.Length = Sepal.Length,
                                Sepal.Width = Sepal.Width,
                                Petal.Length = Petal.Length,
                                Petal.Width = Petal.Width)) +
       geom_serialaxes(stat = "dotProduct") +
       geom_serialaxes_quantile(stat = "dotProduct",
                                quantiles = c(0.25, 0.5, 0.75),
                                colour = c("red", "blue", "green"))
p

scale data

Description

It is mainly used in serial axes

Usage

get_scaledData(
  data,
  sequence = NULL,
  scaling = c("data", "variable", "observation", "none"),
  displayOrder = NULL,
  reserve = FALSE,
  as.data.frame = FALSE
)

Arguments

Deprecated functions in package ggmulti.

Description

The functions listed below are deprecated and will be defunct in the near future. When possible, alternative functions with similar functionality are also mentioned. Help pages for deprecated functions are available at help("coord_radar")

Polygon glyph coordinates

Description

polygon coordinates scaled to (0, 1)

Usage

x_star

y_star

x_cross

y_cross

x_hexagon

y_hexagon

x_airplane

y_airplane

x_maple

y_maple

Format

An object of class numeric of length 10.

An object of class numeric of length 10.

An object of class numeric of length 12.

An object of class numeric of length 12.

An object of class numeric of length 6.

An object of class numeric of length 6.

An object of class numeric of length 32.

An object of class numeric of length 32.

An object of class numeric of length 26.

An object of class numeric of length 26.

See Also

geom_polygon_glyph

Examples

if(require("grid")) {
  library(grid)
  grid.newpage()
  grid.polygon(x=(x_star + 1)/2,
               y=(y_star + 1)/2)
  grid.newpage()
  grid.polygon(x=(x_cross + 1)/2,
               y=(y_cross + 1)/2)
  grid.newpage()
  grid.polygon(x=(x_hexagon + 1)/2,
               y=(y_hexagon + 1)/2)
  grid.newpage()
  grid.polygon(x=(x_airplane + 1)/2,
               y=(y_airplane + 1)/2)
  grid.newpage()
  grid.polygon(x=(x_maple + 1)/2,
               y=(y_maple + 1)/2)
}

Dodge overlapping objects side-to-side

Description

Dodging preserves the vertical position of an geom while adjusting the horizontal position. position_dodge_() dodges bars side by side but conditional on locations.

Usage

position_dodge_(width = NULL, preserve = c("total", "single"))

position_dodge2_(
  width = NULL,
  preserve = c("total", "single"),
  padding = 0.1,
  reverse = FALSE
)

Arguments

Details

It is built based on position_dodge, but used for multiple locations, such as geom_hist_() or geom_density_(). Check examples to see the difference.

See Also

See geom_hist_ and geom_serialaxes_hist for more examples.

Other position adjustments for multiple locations: position_identity_, position_stack_, position_fill_

Parent: position_dodge

Examples

if(require(dplyr)) {
p <- iris %>%
  tidyr::pivot_longer(cols = -Species,
                      names_to = "Outer sterile whorls",
                      values_to = "values") %>%
  ggplot(data,
         mapping = aes(x = `Outer sterile whorls`,
                       y = values,
                       fill = Species))

p + geom_hist_(position = position_dodge_())
}


# all bins are shifted on the left
p +
  geom_hist_(position = position_dodge())

Don't adjust position

Description

Don't adjust position

Usage

position_identity_()

See Also

Other position adjustments for multiple locations: position_stack_, position_fill_, position_dodge_, position_dodge2_

Stack overlapping objects on top of each another

Description

position_stack_ stacks bars on top of each other, conditional on locations.

Usage

position_stack_(vjust = 1, reverse = FALSE)

position_fill_(vjust = 1, reverse = FALSE)

Arguments

Details

It is built based on position_stack, but used for multiple locations, such as geom_hist_ or geom_density_. Rather than stack everything on top of each other, position_stack_ stacks bars based on locations. Check examples to see the difference.

See Also

See geom_hist_, geom_density_, geom_serialaxes_density and geom_serialaxes_hist for more examples.

Other position adjustments for multiple locations: position_identity_, position_dodge_, position_dodge2_

Parent: position_stack

Examples

p <- ggplot(iris,
       mapping = aes(Sepal.Length = Sepal.Length,
                     Sepal.Width = Sepal.Width,
                     Petal.Length = Petal.Length,
                     Petal.Width = Petal.Width,
                     colour = Species))
p +
 geom_serialaxes_density(position = position_stack_())


p +
  geom_serialaxes_density(position = position_stack())