## knitr configuration: http://yihui.name/knitr/options#chunk_options
opts_chunk$set(comment = "", error= TRUE, warning = FALSE, message = FALSE,
tidy = FALSE, cache = F, echo = T,
fig.width = 6, fig.height = 6)
## R configuration
options(width = 116, scipen = 5)
library(ggplot2)
data(diamonds)
p <- ggplot(data = diamonds, mapping = aes(x = carat, y = price, color = cut))
summary(p)
data: carat, cut, color, clarity, depth, table, price, x, y, z [53940x10]
mapping: x = carat, y = price, colour = cut
faceting: facet_null()
layer(geom, geom_params, stat, stat_params, data, mapping, position)
p + layer(geom = "point")
point geom comes with identity stat and identitiy position
geom_point(mapping = NULL, data = NULL,
stat = "identity", position = "identity",
na.rm = FALSE, ...)
More explicit expression
p + layer(geom = "point", stat = "identity", position = "identity")
Abbreviated expression
p + geom_point()
Map carat to x axis for histogram.
p <- ggplot(data = diamonds, mapping = aes(x = carat))
summary(p)
data: carat, cut, color, clarity, depth, table, price, x, y, z [53940x10]
mapping: x = carat
faceting: facet_null()
bar geom and histogram geom come with bin stat and stack position by default
geom_bar(mapping = NULL, data = NULL, stat = "bin",
position = "stack", ...)
geom_histogram(mapping = NULL, data = NULL, stat = "bin",
position = "stack", ...)
Create histogram explicitly
p2 <- p + layer(
geom = "bar",
geom_params = list(fill = "white", color = "steelblue"),
stat = "bin",
stat_params = list(binwidth = 0.5)
)
summary(p2)
data: carat, cut, color, clarity, depth, table, price, x, y, z [53940x10]
mapping: x = carat
faceting: facet_null()
-----------------------------------
geom_bar: fill = white, colour = steelblue
stat_bin: binwidth = 0.5
position_stack: (width = NULL, height = NULL)
p2
Another way of doing the same
p + layer(
geom = "histogram",
geom_params = list(fill = "white", color = "steelblue"),
stat = "bin",
stat_params = list(binwidth = 0.5)
)
Load mammals sleep dataset
data(msleep)
head(msleep)
name genus vore order conservation sleep_total sleep_rem sleep_cycle awake
1 Cheetah Acinonyx carni Carnivora lc 12.1 NA NA 11.9
2 Owl monkey Aotus omni Primates <NA> 17.0 1.8 NA 7.0
3 Mountain beaver Aplodontia herbi Rodentia nt 14.4 2.4 NA 9.6
4 Greater short-tailed shrew Blarina omni Soricomorpha lc 14.9 2.3 0.1333 9.1
5 Cow Bos herbi Artiodactyla domesticated 4.0 0.7 0.6667 20.0
6 Three-toed sloth Bradypus herbi Pilosa <NA> 14.4 2.2 0.7667 9.6
brainwt bodywt
1 NA 50.000
2 0.01550 0.480
3 NA 1.350
4 0.00029 0.019
5 0.42300 600.000
6 NA 3.850
Map REM sleep proportion to x axis, and awake time to y axis
p <- ggplot(data = msleep, mapping = aes(x = sleep_rem / sleep_total, y = awake))
summary(p)
data: name, genus, vore, order, conservation, sleep_total, sleep_rem, sleep_cycle, awake, brainwt,
bodywt [83x11]
mapping: x = sleep_rem/sleep_total, y = awake
faceting: facet_null()
smooth geom comes with smooth stat and vice versa
geom_smooth(mapping = NULL, data = NULL, stat = "smooth",
position = "identity", ...)
stat_smooth(mapping = NULL, data = NULL, geom = "smooth",
position = "identity", method = "auto",
formula = y ~ x, se = TRUE, n = 80, fullrange = FALSE,
level = 0.95, na.rm = FALSE, ...)
Thus, adding geom = “smooth” and stat = “smooth” are the same
p3.a <- p + layer(geom = "point") + layer(geom = "smooth")
p3.b <- p + layer(geom = "point") + layer(stat = "smooth")
summary(p3.a)
data: name, genus, vore, order, conservation, sleep_total, sleep_rem, sleep_cycle, awake, brainwt,
bodywt [83x11]
mapping: x = sleep_rem/sleep_total, y = awake
faceting: facet_null()
-----------------------------------
geom_point:
stat_identity:
position_identity: (width = NULL, height = NULL)
geom_smooth:
stat_smooth:
position_identity: (width = NULL, height = NULL)
p3.a
Give additional arguments to smooth stat via stat_param
p + layer(geom = "point") +
layer(stat = "smooth", stat_params = list(method = "lm", se = F))
Give more additional arguments to smooth geom via geom_params
p + layer(geom = "point") +
layer(geom = "smooth",
geom_params = list(color = "red", alpha = 0.5),
stat = "smooth",
stat_params = list(method = "lm", se = F)
)
A layer can be stored in an object and recycled
smoother <- geom_smooth(color = "red", alpha = 0.5, method = "lm", se = F)
p4 <- p + geom_point() + smoother
p4
ggplot(data = diamonds, mapping = aes(x = carat, y = price, color = cut)) + geom_point() + smoother
The whole structure can be recycled and used with a different data frame
p4
p4 %+% msleep[1:30,]
‘aes’ creates a list of unevaluated expressions. This function
also performs partial name matching, converts color to colour, and
old style R names to ggplot names (eg. pch to shape, cex to size)
Create ggplot object without mapping and then add mapping with aes()
data(mtcars)
p.mtcars <- ggplot(data = mtcars)
summary(p.mtcars)
data: mpg, cyl, disp, hp, drat, wt, qsec, vs, am, gear, carb [32x11]
faceting: facet_null()
p.mtcars <- p.mtcars + aes(x = wt, y = hp)
summary(p.mtcars)
data: mpg, cyl, disp, hp, drat, wt, qsec, vs, am, gear, carb [32x11]
mapping: x = wt, y = hp
faceting: facet_null()
Add additional mapping within layer
p.mtcars <- p.mtcars + layer(geom = "point", mapping = aes(color = factor(cyl)))
summary(p.mtcars)
data: mpg, cyl, disp, hp, drat, wt, qsec, vs, am, gear, carb [32x11]
mapping: x = wt, y = hp
faceting: facet_null()
-----------------------------------
mapping: colour = factor(cyl)
geom_point:
stat_identity:
position_identity: (width = NULL, height = NULL)
Remove mapping by using NULL
p.mtcars <- p.mtcars + aes(y = NULL)
summary(p.mtcars)
data: mpg, cyl, disp, hp, drat, wt, qsec, vs, am, gear, carb [32x11]
mapping: y = NULL, x = wt
faceting: facet_null()
-----------------------------------
mapping: colour = factor(cyl)
geom_point:
stat_identity:
position_identity: (width = NULL, height = NULL)
p.mtcars <- ggplot(data = mtcars, mapping = aes(x = mpg, y = wt))
summary(p.mtcars)
data: mpg, cyl, disp, hp, drat, wt, qsec, vs, am, gear, carb [32x11]
mapping: x = mpg, y = wt
faceting: facet_null()
p.mtcars + geom_point()
Setting color to darkblue (correct)
p.mtcars + geom_point(color = "darkblue")
Mapping color to a categorical variable with value “darkblue” (wrong)
p.mtcars + geom_point(mapping = aes(color = "darkblue"))
Load longitudinal data Oxboys
library(nlme)
data(Oxboys)
head(Oxboys)
Grouped Data: height ~ age | Subject
Subject age height Occasion
1 1 -1.0000 140.5 1
2 1 -0.7479 143.4 2
3 1 -0.4630 144.8 3
4 1 -0.1643 147.1 4
5 1 -0.0027 147.7 5
6 1 0.2466 150.2 6
p.ox <- ggplot(data = Oxboys, mapping = aes(x = age, y = height, group = Subject))
summary(p.ox)
data: Subject, age, height, Occasion [234x4]
mapping: x = age, y = height, group = Subject
faceting: facet_null()
Plot with appropriate grouping
p.ox <- p.ox + layer(geom = "line")
p.ox
Plot without grouping
p.ox + aes(group = NULL)
Subject level grouping for line but no grouping for smooth
p.ox +
layer(geom = "smooth",
geom_params = list(lwd = 3),
stat_params = list(se = F, method = "lm"),
mapping = aes(group = NULL))
geom_boxplot(mapping = NULL, data = NULL,
stat = "boxplot", position = "dodge",
outlier.colour = "black", outlier.shape = 16,
outlier.size = 2, notch = FALSE, notchwidth = 0.5, ...)
stat_boxplot(mapping = NULL, data = NULL,
geom = "boxplot", position = "dodge", na.rm = FALSE,
coef = 1.5, ...)
p.ox2 <- ggplot(data = Oxboys, mapping = aes(x = Occasion, y = height))
summary(p.ox2)
data: Subject, age, height, Occasion [234x4]
mapping: x = Occasion, y = height
faceting: facet_null()
p.ox3 <- p.ox2 + geom_boxplot() + geom_line(mapping = aes(group = Subject), color = "#3366FF")
summary(p.ox3)
data: Subject, age, height, Occasion [234x4]
mapping: x = Occasion, y = height
faceting: facet_null()
-----------------------------------
geom_boxplot: outlier.colour = black, outlier.shape = 16, outlier.size = 2, notch = FALSE, notchwidth = 0.5
stat_boxplot:
position_dodge: (width = NULL, height = NULL)
mapping: group = Subject
geom_line: colour = #3366FF
stat_identity:
position_identity: (width = NULL, height = NULL)
p.ox3
‘stat_summary’ allows for tremendous flexibilty in the
specification of summary functions. The summary function can
either operate on a data frame (with argument name ‘fun.data’) or
on a vector (‘fun.y’, ‘fun.ymax’, ‘fun.ymin’).
stat_summary(mapping = NULL, data = NULL,
geom = "pointrange", position = "identity", ...)
summary stat and line geom do not work together (bug?)
p.ox3 + layer(geom = "line",
geom_params = list(color = "red", mapping = aes(group = 1)),
stat = "summary",
stat_params = list(fun.y = mean))
geom_path: Each group consist of only one observation. Do you need to adjust the group aesthetic?
Create summary statistics data frame and override data option
library(plyr)
mean.at.occasions <-
ddply(.data = Oxboys, .variables = "Occasion",
.fun = summarize, mean.age = mean(age), mean.height = mean(height))
p.ox3 + layer(data = mean.at.occasions,
geom = "line",
geom_params = list(color = "red", size = 2),
stat = "identity",
mapping = aes(y = mean.height, group = 1))
Further summary statistics outside ggplot2
Red lines for min, mean, and max
height.at.occasions <-
ddply(.data = Oxboys, .variables = "Occasion",
.fun = summarize,
mean.height = mean(height),
min.height = min(height),
max.height = max(height))
library(reshape2)
height.at.occasions.melt <- melt(height.at.occasions)
p.ox3 + layer(data = height.at.occasions.melt,
geom = "line",
geom_params = list(color = "red", size = 2),
stat = "identity",
mapping = aes(y = value, group = variable))
Add quantile regression lines
X axis has to be numeric.
p.ox3 + layer(geom = "quantile",
geom_params = list(col = "red", size = 3),
stat = "quantile",
stat_params = list(quantiles = c(0, 0.25, 0.5, 0.75, 1)),
mapping = aes(x = as.numeric(Occasion)))
bin stat used by histogram geom creates variables shown below.
count: number of points in bin
density: density of points in bin, scaled to integrate to 1
ncount: count, scaled to maximum of 1
ndensity: density, scaled to maximum of 1
Map variables newly created by bin stat to y axis
ggplot(data = diamonds, mapping = aes(x = carat)) +
layer(geom = "histogram",
stat = "bin",
mapping = aes(y = ..count..))
ggplot(data = diamonds, mapping = aes(x = carat)) +
layer(geom = "histogram",
stat = "bin",
mapping = aes(y = ..density..))
To avoid overlaps, use position.
p.pos <- ggplot(data = diamonds, mapping = aes(x = clarity, fill = cut))
## stacking
p.pos + layer(geom = "bar", position = "stack")
## filling
p.pos + layer(geom = "bar", position = "fill")
## dodging
p.pos + layer(geom = "bar", position = "dodge")
d <- ggplot(data = diamonds, mapping = aes(x = carat)) +
scale_x_continuous(limit = c(0,3))
bin stat and area geom
d + layer(stat = "bin",
geom = "area",
stat_params = list(binwidth = 0.1))
d + layer(stat = "bin",
geom = "area",
stat_params = list(binwidth = 0.1),
mapping = aes(fill = cut))
bin stat and point geom
d + layer(stat = "bin",
geom = "point",
stat_params = list(binwidth = 0.1),
mapping = aes(y = ..count.., size = ..density..))
bin stat and tile geom
d + layer(stat = "bin",
geom = "tile",
stat_params = list(binwidth = 0.1),
mapping = aes(y = 1, fill = ..count..))
Fit mixed model
library(nlme)
res.lme <- lme(height ~ age, data = Oxboys, random = ~ 1 + age | Subject)
Create plot of raw data
oplot <- ggplot(data = Oxboys, mapping = aes(x = age, y = height, group = Subject)) + geom_line()
Perform prediction and plot
preds <- expand.grid(age = seq(from = -1, to = 1, length = 10),
Subject = levels(Oxboys$Subject))
preds$height <- predict(object = res.lme, newdata = preds)
oplot + geom_line(data = preds, color = "#3366FF", size = 0.4)
Create fitted values and residuals
Oxboys$fitted <- predict(res.lme)
Oxboys$resid <- with(Oxboys, fitted - height)
Plot residuals for each person longitudinally with overall smoother
oplot %+% Oxboys + aes(y = resid) + geom_smooth(aes(group = 1))
Refit a model with age squared term and plot residuals over logitudinally
res.lme2 <- update(res.lme, height ~ age + I(age ^ 2))
Oxboys$fitted2 <- predict(res.lme2)
Oxboys$resid2 <- with(Oxboys, fitted2 - height)
oplot %+% Oxboys + aes(y = resid2) + geom_smooth(aes(group = 1))
