eroicaleo.github.io

Lecture 1

02 Principles of Analytic Graphics

Always show comparison, compare evidence between 2 different hypothesis.
Show causality, mechanism, explanation.
Show multivariate data, some variables are confounding variables.
Integration evidence, different modes of evidence.
Describe and document the appropriate labels.
Content is King.

03 Exploratory Graphs

Why we need graphs?

We want to understand the data
We want to find patterns
To suggest modeling strategies
To debug analysis

Characteristics of exploratory graphs

made quickly
made in large numbers
goal is for personal understandings

Simple Summary of data

Five number of summary
```
 summary(pollution$pm25)
```

boxplot

 boxplot(pollution$pm25, col = "blue")
 # We can also overlay features
 # This will draw a line y = 12
 abline(h = 12)

histogram

 hist(pollution$pm25, col = "green")
 hist(pollution$pm25, col = "green", breaks = 100)
 # density plot, add a strip under the histogram indicating location of each data point
 rug(pollution$pm25)
 # This will draw a line x = 12, with width 2.
 abline(v = 12, lwd = 2)
 # This will draw a line cross the median data point.
 abline(v = median(pollution$pm25), col = "magenta", lwd = 4)

barplot: it is for categorical data.

 barplot(table(pollution$region), col = "wheat", main = "Number of Counties in Each Region")

boxplot for 2 dimensional data

 boxplot(pm25 ~ region, data = pollution, col = "red")

histogram for 2 dimensional data

 par(mfrow = c(1, 2), mar = c(4, 4, 2, 1))
 hist(subset(pollution, region == "east")$pm25, col = "green")
 hist(subset(pollution, region == "west")$pm25, col = "green")

mar is for margin, see the man page of par.

scatterplot

 with(pollution, plot(latitude, pm25))
 with(pollution, plot(latitude, pm25, col = region))

From the above examples, we can see there is two ways to refer to the data in a dataframe.

plot(y ~ x, data = pollution) The ~ is called formula notation.
with(pollution, plot(y, x))

Further resource

R graph gallery

R Bloggers

04 Plotting system in R

base plot system “Artist’s palette” model, add pieces one by one. generation, plot or like function. Then annotation, add labels, axis etc.
Lattice plot system plots are create at once. Good for condition plot, panel plot. Good for putting many many plots on a screen.
ggplot system It mixes the ideas of both

05 Base Plotting system

The Process of Making a plot

Some questions to think about?

The plot will be on a paper? Screen?
Will the plot be used in screen? web browser? presentation?
Large data go into the plot?
Need to dynamically resize the plot?
Which plotting system we will use?

Base Graphics

There are 2 steps:

Initializing a plot
Annotating the plot

plot and hist will launch a graphics device, if there is no one open. plot has lots of arguments, letting you set title, labels. Most of them are documented in the par function man pages.

3 base graph commands: plot, hist, boxplot.

Important parameters:

pch: plotting character (see the man page for function points for details)
lty: line type
lwd: line width
col: color, can be number, string, hex format, colors() function gives a vector of color by name
xlab: x-axis label
ylab: y-axis label

Use par() function to specify these parameters and also to read the current value of them, like par("col"). Don’t forget the double quote.

las: the orientation of axis labels on the plot
bg: background color
mar: margin, start from the bottom and clockwise turn. The unit is line of text.
oma: the outer margin
mfrow: number of plots per row and per column, filled row-wise.
mfcol: number of plots per column and per row, filled column-wise.

Base plotting functions

plot: make a scatterplot, or other plot depending on the class of the objects being plotted.
lines: add lines to a plot.
points: add points to a plot.
text: add text labels to a plot.
title: add titles.
mtext: m means margin, add text to margins.
axis: add axis ticks and labels.
legend: add legend. If they are the line, specify lty. If they are character, specify pch.

Summary

Very flexible and offers high degree of control, but maybe tedious.

06 Base Plotting Demonstration

We can use the examples function to see the examples of a function like examples(points)

pch 21 ~ 25 are similar to 1 ~ 6, but they have boundaries (with col parameters) and fills (with bg parameters).

We could make the plot, but don’t put the data in it by doing plot(x, y, type = "n")

g <- gl(2, 50, labels = c("Male", "Female"))
plot(x, y, type = "n")
plot(x[g == "Male"], y[g == "Male"], col = "blue")
plot(x[g == "Female"], y[g == "Female"], col = "green")

07 Graphic devices

What is a graphic device

A graphic device is something we can make plot apprear
- a window, it’s a screen device
- A PDF file
- A PNG, JPEG file
- A SVG file (scalable vector graphic)
When making a plot, it has to be sent to a graphic device
To open screen device
- quatz() on Mac, X11() on Linux, windows() on Windows
- ?Device to find all devices
plot, xyplot, qplot will send plot to screen device. And there is only one screen device for all the 3 platforms
file devices are appropriate for paper/slide/presentation

How does a plot get created?

Explicitly launch a graphic device

pdf(file = "myplot.pdf")

Call a plot function to make a plot
Annotate the plot
Explicitly close the graphic device by calling dev.off()

Graphic file devices

Vector formats:

Pros: useful for line graphics, resize well, won’t distort.
Cons: not good for plot with lots of points/objects
pdf
svg, useful for animation and web-based plot
windows metafile: windows only
postscript: not use very often

Bitmap formats:

Pros: good for plot with lots of points
Cons: doesn’t resize well
png
jpeg: good for natural scenes, very small, not good for line drawing
tiff
bmp

Multiple open graphic devices

can launch multiple devices
only one is active at a time
each one gets a integer nubmer >= 2
find the current device by dev.cur()
set active device by dev.set(integer)

Copy plots

dev.copy(png, file = "myfile.png")
dev.off()
# if want PDF
dev.copy2pdf

Warning: the plot may not be exactly the same as seen in screen

Lecture 02

Lattice plotting system

Introduction

Lattice contains function to produce trellis graphics, including:
- xyplot
- bwplot
- levelplot
It builds on grid package, which we seldom use directly
Doesn’t have two phases: plotting and annotation
All plotting/annotation is done at once with a single function call

Important functions

xyplot: scatterplot
bwplot: boxplot
histogram: histograms
stripplot: boxplot with actual points
dotplot：plot dots like “violin strings”
splom：scatterplot matrix; like the paris in base system
levelplot, contourplot: for plotting image data

`xyplot` function

xyplot(y ~ x | f * g, data)

Again, we use formula notation, left of ~ is y-axis, right of ~ is x-axis.
f and g are called conditioning variables, which are optional
- they are categorical variables that we condition on
- it means we want to look at the scatterplot of y and x at every level of f and g
- Don’t have to use 2 categorical variables, * indicates interaction.
data is the dataframe
- if no dataframe passed, it will look into parent frame.

Simple lattice plot

Basic one

library(lattice)
library(datasets)
xyplot(Ozone ~ Wind, data = airquality)

Better one, pay attention how we use transform to change the variable in a dataframe

library(lattice)
library(datasets)
xyplot(Ozone ~ Wind, data = airquality)
airquality <- transform(airquality, Month = factor(Month))
xyplot(Ozone ~ Wind | Month, data = airquality, layout = c(5, 1))

Lattice behaviour

Fundamental difference between base plot system

Base system plot to graphic devices.
Lattice system returns a trellis object.
print methods for lattice functions do the plotting work
It’s better to keep the data and code
On the command line, trellis objects are auto-printed.

Lattice panel functions

Lattice functions have panel functions which controls what happens inside each panel
Can supply our own to customize the panel
panel functions receive the x/y coordinates of the data points in their panel
You cannot use the annotation function in base plotting system, you cannot mix the two plotting systems.

# Panel functions
xyplot(y ~ x | f, panel = function(x, y, ...) {
  panel.xyplot(x, y, ...)
  panel.abline(h = median(y), lty = 2)
})

xyplot(y ~ x | f, panel = function(x, y, ...) {
  panel.xyplot(x, y, ...)
  panel.lmline(x, y, col = 2)
})

Summary

Lattice functions are constructed with one single function call
margins and spacing are handled automatically
ideal for creating conditional plots where you examine the same kind of plot under many different conditions
panel functions can be specified and customized to modify what is plotted in each of the plot panels

ggplot2

What is ggplot2?

An implementation of grammar of graphics.
Written by Hadley Wickham
A 3rd graphics system in R
grammar of graphics represents abstraction of graphic ideas and objects
Think “verb”, “noun”, “adjective” for graphics
Allow theory of graphics to build new graph and graphic objects
Shorten the distance from mind to page

Grammar of Graphics

Statistic graph is a mapping from data to aesthetic attributes (color, shape, size) and geometric objects (points, lines, bars). The plot may contain statistic transformation of data and is drawn on a specific coordinate system

The basics `qplot`

much like the plot in base system
must look for a dataframe, if cannot find, look in parent environment.
Plots are made up of aesthetic and geoms
Factors are indicating subsets of data. They should be labeled.
qplot hides underneath
ggplot is core function and very flexible

ggplot2 part2

# installation
install.packages("ggplot2")

Hello word for ggplot2

library(ggplot2)
str(mpg)
qplot(displ, hwy, data = mpg)

Aesthetic

We map the drv variable to different colors, and the plot is automatically labeled.

qplot(displ, hwy, data = mpg, color = drv)

Adding geoms

We can add a smooth line here, note that we want 2 geometric objects here, the data points themselves and the a smooth line.

qplot(displ, hwy, data = mpg, geom = c("point", "smooth"))

Histogram

Make histogram by just specify single variable. Note that here, we need to use fill argument to specify colors.

qplot(hwy, data = mpg, fill = drv)

Like panels in lattice. We want distinguish different subsets of a dataframe. One option is use different color code, another is to use different panels.
The facets argument takes such format, a variable on the left side and a variable on the right side and they are separated by a ~.
The left side is the row of facets and the right side is the column of the facets. If there is nothing to specify, just use .

qplot(displ, hwy, data = mpg, facets = . ~ drv)
qplot(hwy, data = mpg, facets = drv ~ ., binwidth = 2)

Density smooth

qplot(log(eno), data = maacs, geom = "density", color = mopos)

Scatterplot

In addition to separate subset by color code, we can also use shape argument
We can also add smooth line, the default smooth fitting method is “loose”, we can specify by changing the argument method

# separate by shape
qplot(log(eno), log(pm25), data = maacs, shape = mopos)
# separate by color
qplot(log(eno), log(pm25), data = maacs, color = mopos)
# Adding linear regression model smooth line
qplot(log(eno), log(pm25), data = maacs, color = mopos, geom = c("point", "smooth"), method = "lm")
# separate by facets argument
qplot(log(eno), log(pm25), data = maacs, facets = . ~ mopos, geom = c("point", "smooth"), method = "lm")

Summary of qplot

Anolog to plot but with many built-in features
Syntax between base and lattice system
Nice graphics
Don’t bother to customize it, use ggplot2 full power

`ggplot` part3

Basic components

dataframe: the data source
aesthetic mappings: color, size
geometric objects: points, lines, bars, tiles
facets: for conditional graph
stats: statistical transformation: binning, quantiles, smoothing
scales: scale aesthetic mapping uses, e.g. male = red, female = blue
coordinate system

Building Plots with ggplot2

Artist’s palette model
Plots are built in layers
- plot the data
- Overlay a summary
- Metadata and annotation

qplot(logpm25, NocturnalSympt, data = maacs, facets = . ~ bmicat, geom = c("point", "smooth"), method = "lm")

# Initial call to ggplot, specify dataframe, x, y
g <- ggplot(maacs, aes(logpm25, NocturnalSympt))
# Add objects to plot using +
p <- g + geom_point()
print(p)

# Can add smooth line
p <- g + geom_point() + geom_smooth()
p <- g + geom_point() + geom_smooth(method = "lm")
# Then add facets
# The labels are from the variable
# It's better to make sure to label data properly
p <- p + facet_grid(. ~ bmicat)

Annotation

Labels: xlab, ylab, lab, ggtitle
Each geom function has options to modify
For things that make sense globally, use theme()
- theme(legend.position = "none")
Two standard appearance: theme_gray(), theme_bw()

geom_point(color = "steelblue", alpha = 1/2, size = 4)
# Note that if I want to assign color to different data, I have to wrap it in
# aes() function, thus subsetting it with different colors based on factor variable values
geom_point(aes(color = bmicat), alpha = 1/2, size = 4)
# Add labels and title
+labs(title = "MAACS Cohort")
+labs(x = expression("log " * PM[2.5]), y = "Nocturnal Symptoms")
# Modify smooth line, se turns off confidence interval
+ geom_smooth(size = 4, linetype = 3, method = "lm", se = FALSE)
# Change the background and font
+ theme_bw(base_family = "Times")

`ggplot2` part 5

A note about axis limit

Sometimes we may not want to look at the outlier and only focus the typical data

# if we do this, ggplot will subset the data within the range, outlier is excluded
g <- ggplot(testdat, aes(x, y))
g + geom_line() + ylim(-3, 3)
# We might want do
g + geom_line() + coord_cartesian(ylim(-3, 3))

More complex example

We want to see the NO2 and BMI, but NO2 is continous variables. We could use cut() function to make it categorical variable.

Making NO2 Tertile

# Calculate the deciles of the data
cutpoints <- quantile(maacs$logno2_new, seq(0, 1, length = 4), na.rm = TRUE)
# Cut the data at the deciles and create new
maacs$no2dec <- cut(maacs$logno2_new, cutpoints)
# See the levels of new factor variable
levels(maacs$no2dec)

# The real plotting
g <- ggplot(maacs, aes(logpm25, NocturnalSympt))
g + geom_point(alpha = 1/3)
  + facet_wrap(bmicat ~ no2dec, nrow = 3, ncol = 4)
  + geom_smooth(method = "lm", col = "steelblue", se = FALSE)
  + theme_bw(base_family = "Avenir", base_size = 10)
  + labs(x = expression("log " * PM[2.5]))
  + labs(y = "Nocturnal Symptoms")
  + lebs(title = "MAACS Cohort")

Summary of `ggplot`

Very powerful and flexible

eroicaleo.github.io

Lecture 1

02 Principles of Analytic Graphics

03 Exploratory Graphs

04 Plotting system in R

05 Base Plotting system

The Process of Making a plot

Base Graphics

Important parameters:

Base plotting functions

Summary

06 Base Plotting Demonstration

07 Graphic devices

What is a graphic device

How does a plot get created?

Graphic file devices

Multiple open graphic devices

Copy plots

Lecture 02

Lattice plotting system

Introduction

Important functions

xyplot function

Simple lattice plot

Lattice behaviour

Lattice panel functions

Summary

ggplot2

What is ggplot2?

Grammar of Graphics

The basics qplot

ggplot2 part2

Hello word for ggplot2

Aesthetic

Adding geoms

Histogram

Facets

Density smooth

Scatterplot

Summary of qplot

ggplot part3

Basic components

Building Plots with ggplot2

Annotation

ggplot2 part 5

A note about axis limit

More complex example

Making NO2 Tertile

Summary of ggplot

`xyplot` function

The basics `qplot`

`ggplot` part3

`ggplot2` part 5

Summary of `ggplot`