Recap

  • pivot_longer() helps us take our data from wide to long format
    • names_to = gives a new name to the pivoted columns
    • values_to = gives a new name to the values that used to be in those columns
  • pivot_wider() helps us take our data from long to wide format
    • names_from specifies the old column name that contains the new column names
    • values_from specifies the old column name that contains new cell values
  • to merge/join data sets together need a variable in common - usually “id”

📃Cheatsheet

Recap continued

  • to merge/join data sets together need a variable in common - usually “id”
  • ?join - see different types of joining for dplyr
  • inner_join(x, y) - only rows that match for x and y are kept
  • full_join(x, y) - all rows of x and y are kept
  • left_join(x, y) - all rows of x are kept even if not merged with y
  • right_join(x, y) - all rows of y are kept even if not merged with x
  • anti_join(x, y) - all rows from x not in y keeping just columns from x.
  • esquisser() function of the esquisse package can help make plot sketches

📃Cheatsheet

esquisse and ggplot2

esquisse

ggplot2

Why learn ggplot2?

More customization:

  • branding
  • making plots interactive
  • combining plots

Easier plot automation (creating plots in scripts)

Faster (eventually)

ggplot2

ggplot2

Based on the idea of:

layering

plot objects are placed on top of each other with +

📉 +
📈

ggplot2

ggplotcake

Slide Credit: Tanya Shapiro

ggplot2

  • Pros: extremely powerful/flexible – allows combining multiple plot elements together, allows high customization of a look, many resources online

  • Cons: ggplot2-specific “grammar of graphic” of constructing a plot

  • ggplot2 gallery

ggplot with cartoon artists assembling works of art.

Artwork by @allison_horst. https://allisonhorst.com/

Tidy data

To make graphics using ggplot2, our data needs to be in a tidy format

Tidy data:

  1. Each variable forms a column.
  2. Each observation forms a row.

Messy data:

  • Column headers are values, not variable names.
  • Multiple variables are stored in one column.
  • Variables are stored in both rows and columns.

Tidy data: example

Ideally we want each variable as a column and we want each observation in a row.

Column headers are values, not variable names:

Now the the data is “tidy” and in long format

Read more about tidy data and see other examples: Tidy Data tutorial

Data to plot

Type ?Orange for more information.

Is the data in tidy? Is it in long format?

head(Orange)
  Tree  age circumference
1    1  118            30
2    1  484            58
3    1  664            87
4    1 1004           115
5    1 1231           120
6    1 1372           142

First plot with ggplot2 package

First layer of code with ggplot2 package

Will set up the plot - it will be empty!

First layer of code with ggplot2 package

  • Aesthetic mapping aes(x= , y =) describes how variables in our data are mapped to elements of the plot - Note you don’t need to use mapping but it is helpful to know what we are doing.
library(tidyverse) # don't forget to load ggplot2 (part of tidyverse)
# This is not code but shows the general format
ggplot({data_to plot}, aes(x = {var in data to plot},
                                      y = {var in data to plot}))
ggplot(Orange, aes(x = circumference, y = age))

Next layer code with ggplot2 package

There are many to choose from, to list just a few:

  • geom_point() – points (we have seen)
  • geom_line() – lines to connect observations
  • geom_boxplot() – boxplots
  • geom_histogram() – histogram
  • geom_bar() – bar plot
  • geom_col() – column plot
  • geom_tile() – blocks filled with color

Next layer code with ggplot2 package

When to use what plot? A few examples:

  • a scatterplot (geom_point()): to examine the relationship between two sets of continuous numeric data
  • a barplot (geom_bar()): to compare the distribution of a quantitative variable (numeric) between groups or categories
  • a histogram (geom_hist()): to observe the overall distribution of numeric data
  • a boxplot (geom_boxplot()): to compare values between different factor levels or categories

Next layer code with ggplot2 package

Need the + sign to add the next layer to specify the type of plot

ggplot({data_to plot}, aes(x = {var in data to plot},
                                      y = {var in data to plot})) +
  geom_{type of plot}</div>
ggplot(Orange, aes(x = circumference, y = age)) +
  geom_point()

Read as: using Orange data, and provided aesthetic mapping, add points to the plot

Tip - plus sign + must come at end of line

Having the + sign at the beginning of a line will not work!

ggplot(food, aes(x = item_ID,
                           y = item_price_change,
                           fill = item_categ))  
 + geom_boxplot()

Pipes will also not work in place of +!

ggplot(food, aes(x = item_ID,
                           y = item_price_change,
                           fill = item_categ))  %>%
geom_boxplot()

Plots can be assigned as an object

plt1 <- ggplot(Orange, aes(x = circumference, y = age)) +
          geom_point()

plt1

Examples of different geoms

plt1 <- ggplot(Orange, aes(x = circumference, y = age)) +
          geom_point()

plt2 <- ggplot(Orange, aes(x = circumference, y = age)) +
          geom_line()

plt1 # fig.show = "hold" makes plots appear
plt2 # next to one another in the chunk settings

Specifying plot layers: combining multiple layers

Layer a plot on top of another plot with +

ggplot(Orange, aes(x = circumference, y = age)) +
  geom_point() +
  geom_line()

Adding color - can map color to a variable

ggplot(Orange, aes(x = circumference, y = age, color = Tree)) +
  geom_point() +
  geom_line()

Adding color - or change the color of each plot layer

You can change look of each layer separately. Note the arguments like linetype and alpha that allow us to change the opacity of the points and style of the line respectively.

ggplot(Orange, aes(x = circumference, y = age)) +
  geom_point(size = 5, color = "red", alpha = 0.5) +
  geom_line(size = 0.8, color = "black", linetype = 2)

linetype can be given as a number. See the docs for what numbers correspond to what linetype!

Customize the look of the plot

Customize the look of the plot

You can change the look of whole plot using theme_*() functions.

ggplot(Orange, aes(x = circumference, y = age)) +
  geom_point(size = 5, color = "red", alpha = 0.5) +
  geom_line(size = 0.8, color = "brown", linetype = 2) +
  theme_dark()

More themes!

Adding labels

The labs() function can help you add or modify titles on your plot. The title argument specifies the title. The x argument specifies the x axis label. The y argument specifies the y axis label.

ggplot(Orange, aes(x = circumference, y = age)) +
            geom_point(size = 5, color = "red", alpha = 0.5) +
            geom_line(size = 0.8, color = "brown", linetype = 2) +
            labs(title = "My plot of orange tree data",
              x = "Tree Circumference (mm)",
              y = "Tree Age (days since 12/31/1968)")

Adding labels line break

Line breaks can be specified using \n within the labs() function to have a label with multiple lines.

ggplot(Orange, aes(x = circumference, y = age)) +
            geom_point(size = 5, color = "red", alpha = 0.5) +
            geom_line(size = 0.8, color = "brown", linetype = 2) +
            labs(title = "Plot of orange tree data from 1968: \n trunk circumference vs tree age",
              x = "Tree Circumference (mm)",
              y = "Tree Age (days since 12/31/1968)")

Changing axis: specifying axis scale

scale_x_continuous() and scale_y_continuous() can change how the axis is plotted. Can use the breaks argument to specify how you want the axis ticks to be.

range(pull(Orange, circumference))
[1]  30 214
range(pull(Orange, age))
[1]  118 1582
plot_scale <-ggplot(Orange, aes(x = circumference, y = age)) +
                geom_point(size = 5, color = "red", alpha = 0.5) +
                geom_line(size = 0.8, color = "brown", linetype = 2) +
                scale_x_continuous(breaks = seq(from = 20, to = 240, by = 20)) +
                scale_y_continuous(breaks = seq(from = 100, to = 1600, by = 200))

Changing axis: specifying axis scale

plot_scale

ggplot(Orange, aes(x = circumference, y = age)) +
                geom_point(size = 5, color = "red", alpha = 0.5) +
                geom_line(size = 0.8, color = "brown", linetype = 2)

Changing axis: specifying axis limits

xlim() and ylim() can specify the limits for each axis

ggplot(Orange, mapping = aes(x = circumference, y = age)) +
  geom_point(size = 5, color = "red", alpha = 0.5) +
  geom_line(size = 0.8, color = "brown", linetype = 2) +
  labs(title = "My plot of orange tree circumference vs age") +
  xlim(100, max(pull(Orange, circumference)))

Modifying plot objects

You can add to a plot object to make changes! Note that we can save our plots as an object like plt1 below. And now if we reference plt1 again our plot will print out!

plt1 <- ggplot(Orange, aes(x = circumference, y = age)) +
  geom_point(size = 5, color = "red", alpha = 0.5) +
  geom_line(size = 0.8, color = "brown", linetype = 2) +
  labs(title = "My plot of orange tree circumference vs age") +
  xlim(100, max(pull(Orange, circumference)))

plt1 + theme_minimal()

Overwriting specifications

It’s possible to go in and change specifications with newer layers

Orange %>% ggplot(aes(x = circumference,
                      y = age,
                      color = Tree)) +
  geom_line(size = 0.8)

Overwriting specifications

It’s possible to go in and change specifications with newer layers

Orange %>% ggplot(aes(x = circumference,
                      y = age,
                      color = Tree)) +
  geom_line(size = 0.8, color = "black")

GUT CHECK: If we get an empty plot what might we need to do?

A. Add a plot_ layer like plot_point()

B. Add a geom_ layer like geom_point()

GUT CHECK: How do we add more layers in ggplot2 plots?

A. %>%

B. &

C. +

Summary

  • ggplot() specifies what data use and what variables will be mapped to where
  • inside ggplot(), aes(x = , y = , color =) specify what variables correspond to what aspects of the plot in general
  • layers of plots can be combined using the + at the end of lines
  • special theme_*() functions can change the overall look
  • individual layers can be customized using arguments like: size, color alpha (more transparent is closer to 0), and linetype
  • labels can be added with the labs() function and x, y, title arguments - the \n can be used for line breaks
  • xlim() and ylim() can limit or expand the plot area
  • scale_x_continuous() and scale_y_continuous() can modify the scale of the axes
  • by default, ggplot() removes points with missing values from plots.

Lab 1

theme() function:

The theme() function can help you modify various elements of your plot. Here we will adjust the font size of the plot title.

ggplot(Orange, aes(x = circumference, y = age)) +
  geom_point(size = 5, color = "red", alpha = 0.5) +
  geom_line(size = 0.8, color = "brown", linetype = 2) +
  labs(title = "Circumference vs age") +
  theme(plot.title = element_text(size = 20))

theme() function

The theme() function always takes:

  1. an object to change (use ?theme() to see - plot.title, axis.title, axis.ticks etc.)
  2. the aspect you are changing about this: element_text(), element_line(), element_rect(), element_blank()
  3. what you are changing:
    • text: size, color, fill, face, alpha, angle
    • position: "top", "bottom", "right", "left", "none"
    • rectangle: size, color, fill, linetype
    • line: size, color, linetype

theme() function: center title and change size

The theme() function can help you modify various elements of your plot. Here we will adjust the horizontal justification (hjust) of the plot title.

ggplot(Orange, aes(x = circumference, y = age)) +
  geom_point(size = 5, color = "red", alpha = 0.5) +
  geom_line(size = 0.8, color = "brown", linetype = 2) +
  labs(title = "Circumference vs age") +
  theme(plot.title = element_text(hjust = 0.5, size = 20))

theme() function: change title and axis format

ggplot(Orange, aes(x = circumference, y = age)) +
  geom_point(size = 5, color = "red", alpha = 0.5) +
  geom_line(size = 0.8, color = "brown", linetype = 2) +
  labs(title = "Circumference vs age") +
  theme(plot.title = element_text(hjust = 0.5, size = 20),
        axis.title = element_text(size = 16))

theme() function: moving (or removing) legend

If specifying position - use: “top”, “bottom”, “right”, “left”, “none”

ggplot(Orange, aes(x = circumference, y = age, color = Tree)) +
  geom_line()

ggplot(Orange, aes(x = circumference, y = age, color = Tree)) +
  geom_line() +
  theme(legend.position = "none")

Cheatsheet about theme

Keys for specifications

Linetype key

  • geoms that draw lines have a linetype parameter
  • these include values that are strings like “blank”, “solid”, “dashed”, “dotdash”, “longdash”, and “twodash”
Orange %>% ggplot(aes(x = circumference,
                      y = age,
                      color = Tree)) +
  geom_line(size = 0.8, linetype = "twodash")

Keys for specifications

shape key

  • geoms that draw have points have a shape parameter
  • these include numeric values (don’t need quotes for these) and some characters values (need quotes for these)
Orange %>% ggplot(aes(x = circumference,
                      y = age,
                      color = Tree)) +
  geom_point(size = 2, shape = 12)

Can make your own theme to use on plots!

Group and/or color by variable’s values

First, we will read in some data for the purpose of demonstration about food prices over time.

food <- read_csv(
  file = "http://jhudatascience.org/intro_to_r/data/food_price.csv")

Food data

  • 2 different categories (e.g. pasta, rice)
  • 4 different items (e.g. 2 of each category)
  • 10 price values changes collected over time for each item
food
# A tibble: 40 × 4
   item_ID item_categ observation_time item_price_change
   <chr>   <chr>                 <dbl>             <dbl>
 1 ID_1    pasta                     1               1.5
 2 ID_1    pasta                     2               1.5
 3 ID_1    pasta                     3               0.5
 4 ID_1    pasta                     4               2.5
 5 ID_1    pasta                     5               0.5
 6 ID_1    pasta                     6               0.5
 7 ID_1    pasta                     7               1.5
 8 ID_1    pasta                     8               1.5
 9 ID_1    pasta                     9               2.5
10 ID_1    pasta                    10               1.5
# ℹ 30 more rows

Starting a plot

ggplot(food, aes(x = observation_time,
                 y = item_price_change)) +
  geom_line()

If it looks confusing to you, try again

Using group in plots

You can use group element in a mapping to indicate that each item_ID will have a separate price line.

ggplot(food, aes(x = observation_time,
                           y = item_price_change,
                           group = item_ID)) +
  geom_line()

Adding color will automatically group the data

ggplot(food, aes(x = observation_time,
                           y = item_price_change,
                       color = item_ID)) +
  geom_line()

Adding color will automatically group the data

ggplot(food, aes(x = observation_time,
                 y = item_price_change,
                 color = item_categ)) +
  geom_line()

Sometimes you need group and color

ggplot(food, aes(x = observation_time,
                 y = item_price_change,
                 group = item_ID,
                 color = item_categ)) +
  geom_line()

Adding a facet can help make it easier to see what is happening

Two options: facet_grid()- creates a grid shape facet_wrap() -more flexible

Need to specify how you are faceting with the ~ sign.

ggplot(food, aes(x = observation_time,
                 y = item_price_change,
                 color = item_ID)) +
  geom_line() +
  facet_grid( ~ item_categ)

facet_wrap()

  • more flexible - arguments ncol and nrow can specify layout
  • can have different scales for axes using scales = "free_x", scales = "free_y", or scales = "free"
rp_fac_plot <- ggplot(food, aes(x = observation_time, y = item_price_change,color = item_ID)) +
                  geom_line()  +
                  geom_point() +
                  facet_wrap( ~ item_categ, ncol = 1, scales = "free")
rp_fac_plot

Tips!

Let’s talk additional tricks and tips for making ggplots!

Tips - Color vs Fill

  • color is needed for points and lines
  • fill is generally needed for boxes and bars
ggplot(food, aes(x = item_ID,
                 y = item_price_change,
                 color = item_categ)) + #color creates an outline
  geom_boxplot()

ggplot(food, aes(x = item_ID,
                 y = item_price_change,
                 fill = item_categ)) + # fills the boxplot
  geom_boxplot()

Tip - Good idea to add jitter layer to top of box plots

Can add width argument to make the jitter more narrow.

ggplot(food, aes(x = item_ID,
                 y = item_price_change,
                 fill = item_categ)) +
  geom_boxplot() +
  geom_jitter(width = .06)

Tip - be careful about colors for color vision deficiency

scale_fill_viridis_d() for discrete /categorical data scale_fill_viridis_c() for continuous data

ggplot(food, aes(x = item_ID,
                 y = item_price_change,
                 fill = item_categ)) +
  geom_boxplot() +
  geom_jitter(width = .06) +
  scale_fill_viridis_d()

Tip - can pipe data after wrangling into ggplot()

food_bar <-food %>%
  group_by(item_categ) %>%
  summarize("max_price_change" = max(item_price_change)) %>%

ggplot(aes(x = item_categ,
           y = max_price_change,
           fill = item_categ)) +
  scale_fill_viridis_d()+
  geom_col() +
  theme(legend.position = "none")

food_bar

Tip - color outside of aes()

Can be used to add an outline around column/bar plots.

food_bar +
   geom_col(color = "black")

Tip - col vs bar

geom_bar(x =) can only use one aes mapping geom_col(x = , y = ) can have two

Tip - Check what you plot

⚠️ May not be plotting what you think you are! ⚠️

ggplot(food, aes(x = item_ID,
                 y = item_price_change,
                 fill = item_categ)) +
  geom_col()

What did we plot? Always good to check it is correct!

head(food, n = 3)
# A tibble: 3 × 4
  item_ID item_categ observation_time item_price_change
  <chr>   <chr>                 <dbl>             <dbl>
1 ID_1    pasta                     1               1.5
2 ID_1    pasta                     2               1.5
3 ID_1    pasta                     3               0.5
food %>% group_by(item_ID) %>%
  summarize(sum = sum(item_price_change))
# A tibble: 4 × 2
  item_ID   sum
  <chr>   <dbl>
1 ID_1       14
2 ID_2        9
3 ID_3       41
4 ID_4       67

Try that again

food %>% group_by(item_categ, item_ID) %>%
  summarize(mean_change = mean(item_price_change))
# A tibble: 4 × 3
# Groups:   item_categ [2]
  item_categ item_ID mean_change
  <chr>      <chr>         <dbl>
1 pasta      ID_1            1.4
2 pasta      ID_2            0.9
3 rice       ID_3            4.1
4 rice       ID_4            6.7

Try that again

food %>% group_by(item_categ, item_ID) %>%
  summarize(mean_change = mean(item_price_change)) %>%

ggplot(aes(x = item_ID,
           y = mean_change,
           fill = item_categ)) +
  geom_col()

Tip - make sure labels aren’t too small

food_bar +
  theme(text = element_text(size = 20))

Tip- if you need you can remove outliers

Set outlier.shape = NA to get ride of outliers. Be careful about if you really should remove these!

However, if can be helpful if your plot is getting stretched to accommodate plotting an outlier. You can always say in the figure legend what you removed.

esoph1 <- ggplot(esoph, aes(y = ncases, x = agegp)) +
  geom_boxplot()

esoph2 <- ggplot(esoph, aes(y = ncases, x = agegp)) +
  geom_boxplot(outlier.shape = NA)

NA Values

  • if it is a numeric value it will just get dropped from the graph and you will see a warning
  • it is categorical you will see it on the graph and will need to filter to remove the NA category
icecream <-tibble(flavor =
            rep(c("chocolate", "vanilla", NA,"chocolate", "vanilla"), 8))

icecream1 <- ggplot(icecream, aes(x = flavor)) + geom_bar() +
              theme(text=element_text(size=24))

icecream2 <- icecream %>% drop_na(flavor)  %>%
                ggplot( aes(x = flavor)) + geom_bar() +
                  theme(text=element_text(size=24))

Extensions

patchwork package

Great for combining plots together

Also check out the patchwork package

#install.packages("patchwork")
library(patchwork)
(plt1 + plt2)/plt2

directlabels package

plotly

#install.packages("plotly")
library("plotly") # creates interactive plots!
ggplotly(rp_fac_plot)

Also check out the ggiraph package

Saving plots

Saving a ggplot to file

A few options:

  • RStudio > Plots > Export > Save as image / Save as PDF
  • RStudio > Plots > Zoom > [right mouse click on the plot] > Save image as
  • In the code
ggsave(filename = "saved_plot.png",  # will save in working directory
       plot = rp_fac_plot,
       width = 6, height = 3.5)               # by default in inches

GUT CHECK: How to we make sure that the boxplots are filled with color instead of just the outside boarder?

A. Use the fill argument in the aes specification

B. Use color argument in geom_boxplot()

GUT CHECK: If our plot is too complicated to read, what might be a good option to fix this?

A. add more theme() layers

B. Use facet_grid() to split the plot up

Summary

  • The theme() function helps you specify aspects about your plot
    • move or remove a legend with theme(legend.position = "none")
    • change font aspects of individual text elements theme(plot.title = element_text(size = 20))
    • center a title: theme(plot.title = element_text(hjust = 0.5))
  • sometimes you need to add a group element to aes() if your plot looks strange
  • make sure you are plotting what you think you are by checking the numbers!
  • facet_grid(~ variable) and facet_wrap(~variable) can be helpful to quickly split up your plot
    • facet_wrap() allows for a scales = "free" argument so that you can have a different axis scale for different plots
  • use fill to fill in boxplots

Good practices for plots

Check out this guide for more information!

Lab 2

Extra Slides

Customize the look of the plot

You can change the look of whole plot - specific elements, too - like changing and font size - or even more fonts

ggplot(Orange, aes(x = circumference, y = age)) +
  geom_point(size = 5, color = "red", alpha = 0.5) +
  geom_line(size = 0.8, color = "brown", linetype = 2) +
  theme_bw() +
  theme(text=element_text(size=16,  family="Comic Sans MS"))