transitionPlot: A transition plot

Description

This plot's purpose is to illustrate how states change before and after. In my research I use it before surgery and after surgery but it can be used in any situation where you have a change from one state to another

Usage

transitionPlot(
  transition_flow,
  type_of_arrow = c("grid", "simple", "gradient"),
  box_txt = rownames(transition_flow),
  tot_spacing = 0.2,
  box_width = 1/4,
  fill_start_box = "darkgreen",
  txt_start_clr = "white",
  fill_end_box = fill_start_box,
  txt_end_clr = txt_start_clr,
  cex = 2,
  min_lwd = if (type_of_arrow == "grid") 1 else unit(0.1, "mm"),
  max_lwd = if (type_of_arrow == "grid") 6 else unit(5, "mm"),
  lwd_prop_total = TRUE,
  arrow_clr = "#000000",
  abs_arrow_width = FALSE,
  overlap_bg_clr = "#FFFFFF",
  overlap_order = 1:nrow(transition_flow),
  overlap_add_width = if (type_of_arrow == "grid") 1.5 else unit(1, "mm"),
  box_prop,
  mar = unit(rep(3, times = 4), "mm"),
  main = NULL,
  box_label = NULL,
  box_label_pos = "top",
  box_label_cex = cex,
  color_bar = TRUE,
  color_bar_cex = cex * 0.33,
  color_bar_labels,
  color_bar_subspace = NULL,
  new_page = FALSE
)

Value

void

Arguments

transition_flow: This should be a matrix with the size of the transitions. The unit for each cell should be number of observations, row/column-proportions will show incorrect sizes. The matrix needs to be square. The best way to generate this matrix is probably just do a table(starting_state, end_state). The rows represent the starting positions, while the columns the end positions. I.e. the first rows third column is the number of observations that go from the first class to the third class.
type_of_arrow: The types of arrow may be grid, simple, or gradient. Simple grid arrows are the bezierGrob arrows (not that pretty), simple is the bezierArrowSmpl that I've created to get a more exact control of the arrow position and width, while gradient corresponds to bezierArrowGradient allowing the arrow to have a fill color that slowly turns into the color of the arrow.
box_txt: The text to appear inside of the boxes. If you need line breaks then you need to manually add a \n inside the string.
tot_spacing: The proportion of the vertical space that is to be left empty. It is then split evenly between the boxes.
box_width: The width of the box. By default the box is one fourth of the plot width.
fill_start_box: The fill color of the start boxes. This can either be a single value or a vector if you desire different colors for each box. If you specify box_prop then this has to be a 2 column matrix.
txt_start_clr: The text color of the start boxes. This can either be a single value or a vector if you desire different colors for each box. If you specify box_prop then this has to be a 2 column matrix.
fill_end_box: The fill color of the end boxes. This can either be a single value or a vector if you desire different colors for each box. If you specify box_prop then this has to be a 2 column matrix.
txt_end_clr: The text color of the end boxes. This can either be a single value or a vector if you desire different colors for each box. If you specify box_prop then this has to be a 2 column matrix.
cex: The cex gpar of the text
min_lwd: The minimum width of the line that we want to illustrate the tranisition with.
max_lwd: The maximum width of the line that we want to illustrate the tranisition with.
lwd_prop_total: The width of the lines may be proportional to either the other flows from that box, or they may be related to all flows. This is a boolean parameter that is set to true by default, i.e. relating to all flows.
arrow_clr: The color of the arrows. Usually black, can be a vector indicating each arrow from first to last arrow (counting from the top). If the vector is of the same length as the boxes then all box arrows will have the same color (that is all the arrows stemming from the left boxes)
abs_arrow_width: The width can either be absolute, i.e. each arrow headed for a box has the exact same width. The alternative is that the width is related to the line width.
overlap_bg_clr: In order to enhance the 3D perspective and to make it easier to follow arrows the arrows have a background color to separate them from those underneath.
overlap_order: The order from first->last for the lines. This means that the last line will be on top while the first one will appear at the bottom. This should be provided as a vector.
overlap_add_width: The width of the white cross-over line. You can specify this as a scalar multiplication of the current line width. In case of non-grid arrows then you can also have this as a unit which is recommended as it looks better. If the scalar is < 1 then the overlap is ignored.
box_prop: If you want the boxes to have proportions indicating some other factors then input a matrix with quantiles for the proportions. Note the size must be nrow(transition_flow) x 2.
mar: A numerical vector of the form c(bottom, left, top, right) of the type unit()
main: The title of the plot if any, default NULL
box_label: A vector of length 2 if you want to label each box column
box_label_pos: The position of the label, either 'top' or 'bottom'
box_label_cex: The cex of the label, defaults to the default cex
color_bar: If you have proportions inside the transition_flow variable then the color_bar will automatically appear at the bottom unless you set this to FALSE
color_bar_cex: The size of the tick labels for the color bar
color_bar_labels: The labels of the two proportions that make up the color bar. Defaults to the labels of the third dimension for the transition_flow argument.
color_bar_subspace: If there is little or no difference at the low/high proportions of the spectrum then it can be of interest to focus the color change to the center leaving the tails constant
new_page: If you want the plot to appear on a new blank page then set this to TRUE, by default it is FALSE.

Examples

Run this code

# This example does not run since it
# takes a little while to assemble the
# arrows and RMD Check complains that this
# is more than allowed for
library(grid)
par_org <- par(ask = TRUE)
# Settings
no_boxes <- 3
# Generate test setting
transition_matrix <- matrix(NA, nrow = no_boxes, ncol = no_boxes)
transition_matrix[1, ] <- 200 * c(.5, .25, .25)
transition_matrix[2, ] <- 540 * c(.75, .10, .15)
transition_matrix[3, ] <- 340 * c(0, .2, .80)

grid.newpage()
transitionPlot(transition_matrix,
  box_txt = c("First", "Second", "Third"),
  type_of_arrow = "simple",
  min_lwd = unit(1, "mm"),
  max_lwd = unit(6, "mm"),
  overlap_add_width = unit(1, "mm")
)


# Setup proportions
box_prop <- cbind(c(1, 0, 0.5), c(.52, .2, .8))
# From the Set2 Colorbrewer
start_box_clr <- c("#8DA0CB", "#FC8D62")
# Darken the colors slightly
end_box_clr <- c(
  colorRampPalette(c(start_box_clr[1], "#000000"))(10)[2],
  colorRampPalette(c(start_box_clr[2], "#000000"))(10)[2]
)
# Create a new grid
grid.newpage()
transitionPlot(transition_matrix,
  box_prop = box_prop,
  fill_start_box = start_box_clr, fill_end_box = end_box_clr,
  txt_start_clr = c("#FFFFFF", "#000000"), txt_end_clr = c("#FFFFFF", "#000000"),
  box_txt = c("First", "Second", "Third"),
  type_of_arrow = "gradient",
  min_lwd = unit(1, "mm"),
  max_lwd = unit(10, "mm"),
  overlap_add_width = unit(1, "mm")
)
par(par_org)

Run the code above in your browser using DataLab