Sparklines in ggplot2

Question

Tufte Sparklines (as illustrated in his Beautiful Evidence) have been replicated in base graphics as part of YaleToolkit and further perfected as a result of this question. Sparklines have also been done in lattice as a part of my small side project Tufte in R (self-promotion not intended). My goal now is to replicate Tufte sparklines in ggplot2. There are some scripts floating around on Gist and also as a reply to this question on SO, but none of those give a solid base for making replicable sets of sparklines.

Now, I would like those multiple sparklines to look like this (it was done in base graphics and the code is available here) - dots stand for maximum/minimum values, number on right end is a final value in specific time series and grey band shows a rough quantiles range:

I'm not far away but I'm stuck with the assignment of minimal/maximum values and labels:

library(ggplot2)
library(ggthemes)
library(dplyr)
library(reshape)
library(RCurl)
dd <- read.csv(text =
  getURL("https://gist.githubusercontent.com/GeekOnAcid/da022affd36310c96cd4/raw/9c2ac2b033979fcf14a8d9b2e3e390a4bcc6f0e3/us_nr_of_crimes_1960_2014.csv"))
d <- melt(dd, id="Year")
names(d) <- c("Year","Crime.Type","Crime.Rate")
dd <- group_by(d, Crime.Type) %>% 
  mutate(color = (min(Crime.Rate) == Crime.Rate | max(Crime.Rate) == Crime.Rate))
ggplot(dd, aes(x=Year, y=Crime.Rate)) + 
  facet_grid(Crime.Type ~ ., scales = "free_y") + 
  geom_line(size=0.3) + geom_point(aes(color = color)) + 
  scale_color_manual(values = c(NA, "red"), guide=F) +
  theme_tufte(base_size = 15) + 
  theme(axis.title=element_blank(), 
        axis.text.y = element_blank(), axis.ticks = element_blank()) +
  theme(strip.text.y = element_text(angle = 0, vjust=0.2, hjust=0)) 

Diagenesis

Answer 1

Here is one approach to getting single colored points, as well as the three sets of labels and shaded quartile ranges:

# Calculate the min and max values, which.min returns the first (like your example):
mins <- group_by(d, Crime.Type) %>% slice(which.min(Crime.Rate))
maxs <- group_by(d, Crime.Type) %>% slice(which.max(Crime.Rate))
ends <- group_by(d, Crime.Type) %>% filter(Year == max(Year))
quarts <- d %>%
  group_by(Crime.Type) %>%
  summarize(quart1 = quantile(Crime.Rate, 0.25),
            quart2 = quantile(Crime.Rate, 0.75)) %>%
  right_join(d)

ggplot(d, aes(x=Year, y=Crime.Rate)) + 
  facet_grid(Crime.Type ~ ., scales = "free_y") + 
  geom_ribbon(data = quarts, aes(ymin = quart1, ymax = quart2), fill = 'grey90') +
  geom_line(size=0.3) +
  geom_point(data = mins, col = 'blue') +
  geom_text(data = mins, aes(label = Crime.Rate), vjust = -1) +
  geom_point(data = maxs, col = 'red') +
  geom_text(data = maxs, aes(label = Crime.Rate), vjust = 2) +
  geom_text(data = ends, aes(label = Crime.Rate), hjust = 0) +
  geom_text(data = ends, aes(label = Crime.Type), hjust = 0, nudge_x = 5) +
  expand_limits(x = max(d$Year) + (0.25 * (max(d$Year) - min(d$Year)))) +
  scale_x_continuous(breaks = seq(1960, 2010, 10)) +
  scale_y_continuous(expand = c(0.1, 0)) +
  theme_tufte(base_size = 15) +
  theme(axis.title=element_blank(),
        axis.text.y = element_blank(), 
        axis.ticks = element_blank(),
        strip.text = element_blank())

I'm assuming you don't want a legend here. You can almost certainly make things more concise by merging some data.frames, but multiple geom calls seem to be easiest here.

Saboteur

...and the labels could just be two more geom_text layers with vjust set appropriately. – Brusa

Just noticed I swapped the colors... -.- – Saboteur