Visualizing epidemiologic data in R and RStudio

Learning objectives:

1. To put to use the dplyr commands from the first session
2. To make beautiful plots using the ggplot2 package

Life expectancy in the United States by race and gender, 1969-2013

These data are partial results from a study that I did on the difference in life expectancy between non-Hispanic Black and White men and women in the United States over time.

A subset of the results have been stored in the Data/ folder as a CSV file.

Do you remember which function to use to import CSV data into R?

readr’s read_csv() to import these data

library(readr) #readr is part of the tidyverse
le_data <- read_csv("../Data/Life-expectancy-by-state-long.csv")

## Parsed with column specification:
## cols(
##   state = col_character(),
##   stabbrs = col_character(),
##   year = col_double(),
##   sex = col_character(),
##   Census_Region = col_character(),
##   Census_Division = col_character(),
##   LE = col_double(),
##   race = col_character()
## )

Five functions to get to know your dataset

Function 1

head(le_data)

## # A tibble: 6 x 8
##   state   stabbrs  year sex    Census_Region Census_Division       LE race 
##   <chr>   <chr>   <dbl> <chr>  <chr>         <chr>              <dbl> <chr>
## 1 Alabama AL       1969 Female South         East South Central  75.8 white
## 2 Alabama AL       1969 Male   South         East South Central  66.6 white
## 3 Alabama AL       1970 Female South         East South Central  75.9 white
## 4 Alabama AL       1970 Male   South         East South Central  66.7 white
## 5 Alabama AL       1971 Female South         East South Central  76.2 white
## 6 Alabama AL       1971 Male   South         East South Central  66.9 white

Five functions to get to know your dataset

Function 2

dim(le_data)

## [1] 7200    8

Five functions to get to know your dataset

Function 3

names(le_data)

## [1] "state"           "stabbrs"         "year"            "sex"            
## [5] "Census_Region"   "Census_Division" "LE"              "race"

Five functions to get to know your dataset

Function 4

str(le_data)

## Classes 'spec_tbl_df', 'tbl_df', 'tbl' and 'data.frame': 7200 obs. of  8 variables:
##  $ state          : chr  "Alabama" "Alabama" "Alabama" "Alabama" ...
##  $ stabbrs        : chr  "AL" "AL" "AL" "AL" ...
##  $ year           : num  1969 1969 1970 1970 1971 ...
##  $ sex            : chr  "Female" "Male" "Female" "Male" ...
##  $ Census_Region  : chr  "South" "South" "South" "South" ...
##  $ Census_Division: chr  "East South Central" "East South Central" "East South Central" "East South Central" ...
##  $ LE             : num  75.8 66.6 75.9 66.7 76.2 ...
##  $ race           : chr  "white" "white" "white" "white" ...
##  - attr(*, "spec")=
##   .. cols(
##   ..   state = col_character(),
##   ..   stabbrs = col_character(),
##   ..   year = col_double(),
##   ..   sex = col_character(),
##   ..   Census_Region = col_character(),
##   ..   Census_Division = col_character(),
##   ..   LE = col_double(),
##   ..   race = col_character()
##   .. )

Five functions to get to know your dataset

Function 5

View(le_data)

To RStudio!

Summary: Four functions to get to know your dataset

1. head(): prints the first 6 lines of a data frame
2. dim(): prints the # rows and # columns
3. names(): prints the variable names
4. str(): shows the type of each variable and some values
5. View(): opens the viewer pane in RStudio

Life expectancy for White men in California

Make a scatter plot of the life expectancy for White men in California over time.

Since the dataset contains 39 states across two genders and two races, first use a function to subset the data to contain only White men in California.

Which function from Malcolm’s lesson do we need?

dplyr’s filter() to select a subset of rows

library(dplyr)
wm_cali <- le_data %>% filter(state == "California", 
                              sex == "Male", 
                              race == "white")

#this is equivalent:
wm_cali <- le_data %>% filter(state == "California" & sex == "Male" & race == "white")

First step to building a ggplot(): set up a canvas

• The line of code specified the data set and what goes on the x and y axes

library(ggplot2)
ggplot(data = wm_cali, aes(x = year, y = LE))

Second step to building a ggplot(): tell ggplot how to plot the data

ggplot(data = wm_cali, aes(x = year, y = LE)) + geom_point()

• geom_point() tells ggplot to use points to plot these data

labs() to add a title, a caption, and modify x and y axes titles

ggplot(data = wm_cali, aes(x = year, y = LE)) + geom_point() +
  labs(title = "Life expectancy in White men in California, 1969-2013",
       y = "Life expectancy", 
       x = "Year", 
       caption = "Data from Riddell et al. (2018)")

col controls the color of geom_point()

ggplot(data = wm_cali, aes(x = year, y = LE)) + geom_point(col = "blue") +
  labs(title = "Life expectancy in White men in California, 1969-2013",
       y = "Life expectancy", 
       x = "Year", 
       caption = "Data from Riddell et al. (2018)")

size controls the size of geom_point()

ggplot(data = wm_cali, aes(x = year, y = LE)) + geom_point(col = "blue", size = 4) +
  labs(title = "Life expectancy in White men in California, 1969-2013",
       y = "Life expectancy", 
       x = "Year", 
       caption = "Data from Riddell et al. (2018)")

Line plot rather than scatter plot

What if we wanted to make these data into a line plot instead. What part of the code should change?

ggplot(data = wm_cali, aes(x = year, y = LE)) + 
  geom_point(col = "blue", size = 4) +
  labs(title = "Life expectancy in White men in California, 1969-2013",
       y = "Life expectancy", 
       x = "Year", 
       caption = "Data from Riddell et al. (2018)")

geom_line() to make a line plot

ggplot(data = wm_cali, aes(x = year, y = LE)) + geom_line(col = "blue") +
  labs(title = "Life expectancy in White men in California, 1969-2013",
       y = "Life expectancy", 
       x = "Year", 
       caption = "Data from Riddell et al. (2018)")

Life expectancy for White and Black men in California

What do we need to change to make a separate line for both Black and White men?

First, update the filter()

wbm_cali <- le_data %>% filter(state == "California",
                               sex == "Male")

Look at the previous code and output first:

ggplot(data = wm_cali, aes(x = year, y = LE)) + geom_line(col = "blue") +
  labs(title = "Life expectancy in White men in California, 1969-2013",
       y = "Life expectancy", 
       x = "Year", 
       caption = "Data from Riddell et al. (2018)")

And change it to link color to race

ggplot(data = wbm_cali, aes(x = year, y = LE)) + geom_line(aes(col = race)) +
  labs(title = "Life expectancy in Black and White men in California, 1969-2013",
       y = "Life expectancy", 
       x = "Year", 
       caption = "Data from Riddell et al. (2018)")

Always use the aes() function to link a plot feature to a variable in your data frame

The operative word is link. Whenever you want to link something about how the plot looks to a variable in the data frame, you need to link these items inside the aes() function:

ggplot(data = wbm_cali, aes(x = year, y = LE)) + geom_line(aes(col = race)) +
  labs(title = "Life expectancy in Black and White men in California, 1969-2013",
       y = "Life expectancy", 
       x = "Year", 
       caption = "Data from Riddell et al. (2018)")

The aes() function

• What else was added to the plot when you used the aes() function?

The aes() function

• What else was added to the plot when you used the aes() function?
  • A legend was added showing the link between the line color and the data frame’s race variable

What if we also wanted to look at women?

What if we also wanted to look at women?

cali_data <- le_data %>% filter(state == "California")

What is wrong with this plot?

ggplot(data = cali_data, aes(x = year, y = LE)) + geom_line(aes(col = race)) +
  labs(title = "Life expectancy in California, 1969-2013",
       y = "Life expectancy", 
       x = "Year", 
       caption = "Data from Riddell et al. (2018)")

Use lty() to link line type to sex

ggplot(data = cali_data, aes(x = year, y = LE)) + geom_line(aes(col = race, lty = sex)) +
  labs(title = "Life expectancy in California, 1969-2013",
       y = "Life expectancy", 
       x = "Year", 
       caption = "Data from Riddell et al. (2018)")

Use facet_wrap() to make separate plots for a specified variable

ggplot(data = cali_data, aes(x = year, y = LE)) + 
  geom_line(aes(col = race, lty = sex)) +
  labs(title = "Life expectancy in California, 1969-2013",
       y = "Life expectancy", 
       x = "Year", 
       caption = "Data from Riddell et al. (2018)") +
  facet_wrap(~ sex) 

Compare two states

How do we update the filter to include data from California and New York?

Compare two states

updated_data <- le_data %>% filter(state %in% c("California", "New York"))

Let’s write the code together

#to fill in during class

Let’s write the code together

ggplot(data = updated_data, aes(x = year, y = LE)) + 
  geom_line(aes(col = race, lty = sex)) +
  labs(title = "Life expectancy in California and New York, 1969-2013",
       y = "Life expectancy", 
       x = "Year", 
       caption = "Data from Riddell et al. (2018)") +
  facet_grid(state ~ sex)

Question

What is the difference between facet_wrap() and facet_grid()?

So far

• geom_point() to make scatter plots
• geom_line() to make line plots
• col = "blue", size = 2, lty = 2, to change color, size and line type of the geom
• aes(col = race) to link color to race
• aes(lty = sex) to link line type to sex
• facet_wrap(~ var1) to make separate plots for different levels of one variable
• facet_grid(var1 ~ var2) to make separate plots for combinations of levels of two variables

What if we wanted to make a histogram…

…of life expectancy of White men in 2013?

Before you code, try and visualize what the histogram will show

• What is on the x axis?
• What is on the y axis?

Update the filter

wm_data <- le_data %>% filter(year == 2013, sex == "Male", race == "white")

geom_histogram() to make histograms

ggplot(dat = wm_data, aes(x = LE)) + geom_histogram()

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

use fill to change the fill of the histogram and binwidth to specify the bin width

wm_data <- le_data %>% filter(year == 2013, sex == "Male", race == "white")

ggplot(dat = wm_data, aes(x = LE)) + 
  geom_histogram(binwidth = 1, col = "white", fill = "forest green") 

Apply some of our new skills

data_2013 <- le_data %>% filter(year == 2013)

ggplot(dat = data_2013, aes(x = LE)) + 
  geom_histogram(binwidth = 1, col = "white", aes(fill = sex)) + 
  facet_grid(race ~ sex)

Recap: What functions did we learn?

1. ggplot()
   • geom_scatter()
   • geom_line()
   • geom_histogram()
   • aes() to link aesthetics to variables in our data frame
   • facet_wrap(~ var1), facet_grid(var1 ~ var2)
   • labs(title = "Main", y = "y axis", x = "x axis", caption = "below plot")

Recap: What arguments were useful?

2. ggplot()
   • col
   • size
   • lty

We only skimmed the surface!

• You now have a sense of how ggplot works, but you might be itching to learn more.
  • Kieran Healy’s data visualization book
  • RStudio ggplot2 cheatsheet

Where to ask ggplot2 questions

• The RStudio community page
• Stack Overflow
• On Twitter using the #rstats hashtag