David Selby Selbosh

The original graphic (reproduced below) features two concentric doughtnut charts with an accompanying legend, and the reader is challenged to match up eight seemingly arbitrary colours with the different categories.

Proportions are difficult to ascertain because there are too many groups and many of them have similar sizes. All the numbers are printed on the chart which sort of makes it into a badly-laid-out table.

My makeover simplifies everything into a grouped bar chart, with direct labelling of categories and a simple legend to distinguish online from direct services. From this graph it is easy to see that the majority of Citizens Advice issues are dealt with online, whereas inferring this from the original graph requires reading off the actual numbers.

This graphic could be improved further by cleaning up

	# Faster set operations for integer sequences
	import Base.setdiff, Base.union, Base.collect

	function setdiff(a::UnitRange{Int}, b::UnitRange{Int})
	if intersect(a, b) == a
	last(a):first(a)
	elseif isempty(intersect(a, b))
	a
	# a0 --- b0 --- a1 --- b1
	elseif first(a) <= first(b) && last(a) <= last(b)

	#############################
	# Fix a rainbow-coloured plot
	#############################
	rainbow2viridis <- function(img) {
	# Input: `img`, an H x W x 3 array of RGB values from an image file
	dimnames(img) <- list(height = NULL, width = NULL,
	rgb = c('r', 'g', 'b'))
	data <- reshape2::melt(img)
	data <- reshape2::dcast(data, width + height ~ rgb)
	data <- subset(data, !(r == 1 & g == 1 & b == 1)) # omit white pixels

	simulate <- function(means, ..., sampler = rpois) {
	# Simulate from a dataset and return
	# a sample that is in the same format
	sample <- means
	sample[] <- sampler(length(means), means, ...)
	return(sample)
	}

	make_QS <- function(n, mean = 10) {
	D <- diag(sample.int(n))

	# devtools::install_github("ikosmidis/cranly")
	library(cranly)
	package_db <- clean_CRAN_db()
	(cranly_ts <- attr(package_db, "timestamp"))
	package_network <- build_network(package_db)

	library(igraph)
	cranlig <- as.igraph(package_network, reverse = TRUE)
	npkgs <- vcount(cranlig)
	nlinks <- ecount(cranlig)

	# Create data frame with a list column
	users <- data.frame(
	id = 1:3,
	username = c('Tom', 'Dick', 'Harry'),
	following = I(list(
	c('TeaStats', 'RobertJBlincoe'),
	c('JeremyCorbyn', 'realDonaldTrump'),
	c('Scottish_Tweets')
	))
	)

	library(rlang)
	library(dplyr)

	# ----------------------------------------------------------------
	# First implementation: specify desired variable names explicitly.
	# ----------------------------------------------------------------

	test1 <- function(var1, var2, data, name1, name2) {
	qvar1 <- enquo(var1)
	qvar2 <- enquo(var2)

	@import url(https://fonts.googleapis.com/css?family=Merriweather:400,400i,700\|Merriweather+Sans:300,400);
	@import url(https://fonts.googleapis.com/css?family=Source+Code+Pro:400,700);

	body {
	font-family: Merriweather, Georgia, serif;
	color: #515151;
	line-height: 1.5;
	}

	.remark-slide-content {

	license: gpl-3.0
	height: 700
	scrolling: no
	border: no