21 Polishing

Stat405 Polishing, continued

Hadley Wickham
Monday, 9 November 2009

1. Coming up next.
2. Project.
3. Colour scales
4. Themes.


Monday Wednesday

Nov 9 Polishing Spam

Nov 16 Posters / Project Spam

Nov 23 Debugging Thanksgiving
Professional
Nov 30 development
Poster pres


Project 3

Your turn to ﬁnd an interesting data
source and analyse it. Any data is
acceptable but it needs to be large
(>1,000 points).
If you don’t have any ideas, try looking at
http://delicious.com/hadley/data


Deliverables
Poster.
Single double-sided page handout,
summarising the major ﬁndings on the
poster.
The R code used to clean the data and
produce the plots. (No manual steps!)


Timeline
Nov 17-19. Meet to discuss choice of
data and initial analysis ideas. Data
preparation.
Nov 23-30. Review of draft poster
(optional, but highly recommended).
Dec 2. Final poster presentation. Turn in
electronic copy of handout & code.


Polishing


Tools

Scales. Used to override default
perceptual mappings, and tune
parameters of axes and legends.
Themes: control presentation of
non-data elements.


Your turn

Recall the four key parameters that all
scales have in common. What are their
names and what do they do?


p <- qplot(carat, price, data = diamonds, geom = "hex")

# First argument (name) controls legend title
p + scale_fill_continuous("Count")

# Breaks and labels control legend keys
p + scale_fill_continuous(breaks = c(1000, 3500, 7000))
p + scale_fill_continuous(breaks = c(0, 4000, 8000))

# Why don't 0 and 8000 have colours?
p + scale_fill_continuous(breaks = c(0, 4000, 8000),
limits = c(0, 8000))

# Can use labels to make more human readable
breaks <- c(0, 2000, 4000, 6000, 8000)
labels <- format(breaks, big.mark = ",")
p + scale_fill_continuous(breaks = breaks, labels = labels,
limits = c(0, 8000))


Default colour scales

Discrete: evenly spaced hues of equal
chroma and luminance. No colour
appears more important than any other.
Does not imply order.
Continuous: evenly spaced hues
between two colours.


Alternatives

Discrete: hue, brewer, manual
Continuous: gradient, gradient2,
gradientn


Color brewer
Cynthia Brewer applied the basic
principles and then rigorously tested the
results to produce a selection of good
palettes, particularly tailored for maps:
http://colorbrewer2.org/
Can use cut_interval() or cut_number()
to convert continuous to discrete.


# Fancy looking trigonometric function
vals <- seq(-4 * pi, 4 * pi, len = 50)
df <- expand.grid(x = vals, y = vals)
df$r <- with(df, sqrt(x ^ 2 + y ^ 2))
df$z <- with(df, cos(r ^ 2) * exp(- r / 6))
df$z_cut <- cut_interval(df$z, 9)

(p1 <- qplot(x, y, data = df, fill = z,
geom = "tile"))
(p2 <- qplot(x, y, data = df, fill = z_cut,
geom = "tile"))


p1 + scale_fill_gradient(low = "white",
high = "black")

# Highlight deviations
p1 + scale_fill_gradient2()
p1 + scale_fill_gradient2(breaks = seq(-1, 1,
by = 0.25), limits = c(-1, 1))
p1 + scale_fill_gradient2(mid = "white",
low = "black", high = "black")

p2 + scale_fill_brewer(pal = "Blues")


Your turn

Read through the examples for
scale_colour_brewer,
scale_colour_gradient2 and
scale_colour_gradientn.
Experiment!


Colour blindness

7-10% of men are red-green colour
“blind”. (Many other rarer types of colour
blindness)
Solutions: avoid red-green contrasts; use
redundant mappings; test. I like color
oracle: http://colororacle.cartography.ch


Themes


Visual appearance
So far have only discussed how to get the
data displayed the way you want,
focussing on the essence of the plot.
Themes give you a huge amount of
control over the appearance of the plot,
the choice of background colours, fonts
and so on.


# Two built in themes. The default:
qplot(carat, price, data = diamonds)

# And a theme with a white background:
qplot(carat, price, data = diamonds) + theme_bw()

# Use theme_set if you want it to apply to every
# future plot.
theme_set(theme_bw())

# This is the best way of seeing all the default
# options
theme_bw()
theme_grey()


Elements
You can also make your own theme, or
modify and existing.
Themes are made up of elements which
can be one of: theme_line, theme_segment,
theme_text, theme_rect, theme_blank
Gives you a lot of control over plot
appearance.


Elements
Axis: axis.line, axis.text.x, axis.text.y,
axis.ticks, axis.title.x, axis.title.y
Legend: legend.background, legend.key,
legend.text, legend.title
Panel: panel.background, panel.border,
panel.grid.major, panel.grid.minor
Strip: strip.background, strip.text.x,
strip.text.y


p <- qplot(displ, hwy, data = mpg) +
opts(title = "Bigger engines are less efficient")

# To modify a plot
p
p + opts(plot.title =
theme_text(size = 12, face = "bold"))
p + opts(plot.title = theme_text(colour = "red"))
p + opts(plot.title = theme_text(angle = 45))
p + opts(plot.title = theme_text(hjust = 1))


Your turn
Fix the overlapping y labels on this plot:
qplot(reorder(model, hwy), hwy, data =
mpg)
Rotate the labels on these strips so they
are easier to read.
qplot(hwy, reorder(model, hwy), data =
mpg) + facet_grid(manufacturer ~ .,
scales = "free", space = "free")


21 Polishing

More Related Content

Viewers also liked

Similar to 21 Polishing

More from Hadley Wickham

Recently uploaded

21 Polishing