issue commands to R using the Rstudio REPL interface
load a package into R
read some tabluar data into R
visualize tabluar data using ggplot geoms, aesthetics, and facets
Basics
Console
The R console window is the left (or lower-left) window in RStudio. The R console uses a “read, eval, print” loop. This is sometimes called a REPL.
Read: R reads what you type …
Eval: R evaluates it …
Print: R prints the result …
Loop: (repeat forever)
The box contains an expression that will be evaluated by R, followed by the result of evaluating that expression.
1+2
[1] 3
3 is the answer
Ignore the [1] for now.
R performs operations (called functions) on data and values
These can be composed arbitrarily
log(1+3)
[1] 1.386294
paste("The answer is", log(1+3))
[1] "The answer is 1.38629436111989"
How do I…
typing ?function_name gives you information about what the function does
Google is your friend. Try “function_name R language” or “how do I X in R?”. I also strongly recommend using “tidyverse” in your queries or the name of a tidyverse package (more in a moment) that has related functions
stackoverflow is your friend. It might take some scrolling, but you will eventually find what you need
LLMs…
Why learn this when LLMs can do it?
Anthropic’s Claude code (using eg Opus 5) or OpenAI’s codex (using eg Sol) can do everything you will learn about in this class and probably also in any quant class you’ll take in your MPH.
Why learn this when LLMs can do it?
Discuss in your groups.
Quadratic Equation
Solutions to a polynomial equation \(ax^2 + bx + c = 0\) are given by \[x = \frac{-b \pm \sqrt{b^2 - 4ac}}{2a}\]
Figure out how to use R functions and operations for square roots, exponentiation, and multiplication to calculate x given a=3, b=14, c=-5 (no LLMs).
How did this feel? What was your emotional reaction when you saw the question?
What did you learn? What did you notice?
Packages
The amazing thing about programming is that you are not limited to what is built into the language
Millions of R users have written their own functions that you can use
These are bundled together into packages
To use functions that aren’t built into the “base” language, you have to tell R to first go download the relevant code, and then to load it in the current session
install.packages("QuadRoot")
library(QuadRoot)QuadRoot(c(3,14,-5))
[1] "The two x-intercepts for the quadratic equation are 0.3333 and -5.0000."
The tidyverse package has a function called read_csv() that lets you read csv (comma-separated values) files into R.
csv is a common format for data to come in, and it’s easy to export csv files from microsoft excel, for instance.
# I have a file called "nhanes_subset.csv" on github that we can read from the URLnhanes =read_csv("https://raw.githubusercontent.com/alejandroschuler/r4ds-courses/mph-nhanes/data/nhanes/nhanes_subset.csv")
Error in read_csv("https://raw.githubusercontent.com/alejandroschuler/r4ds-courses/mph-nhanes/data/nhanes/nhanes_subset.csv"): could not find function "read_csv"
This fails because I haven’t yet installed and loaded the tidyverse package
install.packages("tidyverse") # go download the package called "tidyverse"- only have to do this oncelibrary("tidyverse") # load the package into the current R session - do this every time you use R and need functions from this package
If your data is not already in csv format, google “covert X format to csv” or “read X format data in R”
We’ll learn the details of this later, but this is enough to get you started!
Looking at data
nhanes is now a dataset loaded into R. To look at it, just type
nhanes
# A tibble: 59 x 11
id age sex race diabetes glucose hba1c bmi waist_cm bp_sys_1
<chr> <dbl> <chr> <chr> <chr> <dbl> <dbl> <dbl> <dbl> <dbl>
1 NHANES-1001~ 72 male Mexi~ Yes 124 6.8 31.5 107. 124
2 NHANES-1001~ 80 fema~ White Yes 95 6.7 26 87 180
3 NHANES-1002~ 66 male White Yes 171 7.9 26.1 105. 120
4 NHANES-1002~ 65 fema~ Black Yes 157 7.4 34.4 102. 102
5 NHANES-1004~ 31 male White No 98 4.8 23.2 85.3 112
6 NHANES-1008~ 80 male Mexi~ Yes 174 7.7 32.1 120. 156
7 NHANES-1011~ 49 fema~ Black No 107 5.7 39.3 112. 128
8 NHANES-1012~ 65 fema~ Black Yes 150 8.1 30.8 114. 128
9 NHANES-1013~ 80 fema~ White Yes 165 8.7 30 118. 156
10 NHANES-1014~ 53 male Asian No 104 5.6 28.6 99.1 126
# i 49 more rows
# i 1 more variable: cholesterol <dbl>
This is a data frame, one of the most powerful features in R (a “tibble” is a kind of data frame). - Similar to an Excel spreadsheet. - One row ~ one instance of some (real-world) object. - One column ~ one variable, containing the values for the corresponding instances. - All the values in one column should be of the same type (a number, a category, text, etc.), but different columns can be of different types.
The Dataset
nhanes
# A tibble: 59 x 11
id age sex race diabetes glucose hba1c bmi waist_cm bp_sys_1
<chr> <dbl> <chr> <chr> <chr> <dbl> <dbl> <dbl> <dbl> <dbl>
1 NHANES-1001~ 72 male Mexi~ Yes 124 6.8 31.5 107. 124
2 NHANES-1001~ 80 fema~ White Yes 95 6.7 26 87 180
3 NHANES-1002~ 66 male White Yes 171 7.9 26.1 105. 120
4 NHANES-1002~ 65 fema~ Black Yes 157 7.4 34.4 102. 102
5 NHANES-1004~ 31 male White No 98 4.8 23.2 85.3 112
6 NHANES-1008~ 80 male Mexi~ Yes 174 7.7 32.1 120. 156
7 NHANES-1011~ 49 fema~ Black No 107 5.7 39.3 112. 128
8 NHANES-1012~ 65 fema~ Black Yes 150 8.1 30.8 114. 128
9 NHANES-1013~ 80 fema~ White Yes 165 8.7 30 118. 156
10 NHANES-1014~ 53 male Asian No 104 5.6 28.6 99.1 126
# i 49 more rows
# i 1 more variable: cholesterol <dbl>
This is a subset of NHANES, the CDC’s National Health and Nutrition Examination Survey. NHANES sends a mobile clinic around the country, interviews a nationally representative sample of Americans, and gives each of them a physical exam and a blood draw.
30 participants with diagnosed diabetes, 29 without
We have basic demographics as well as several body measurements and lab values.
Let’s see if we can find anything interesting from this already-generated data!
Investigating a relationship
Let’s say we’re curious about the relationship between two lab values, glucose and hba1c.
Can we use R to make a plot of these two variables?
ggplot(nhanes) +geom_point(aes(x = glucose, y = hba1c))
ggplot(dataset) says “start a chart with this dataset”
+ geom_point(...) says “put points on this chart”
aes(x=x_values y=y_values) says “map the values in the column x_values to the x-axis, and map the values in the column y_values to the y-axis” (aes is short for aesthetic)
ggplot
ggplot(nhanes) +geom_point(aes(x = waist_cm, y = bmi))
ggplot is short for “grammar of graphics plot”
This is a language for describing how data get linked to visual elements
ggplot() and geom_point() are functions imported from the ggplot2 package, which is one of the “sub-packages” of the tidyverse package we loaded earlier
Exercise: comparing lab values
Make a scatterplot of diabetes vs hba1c (another lab value in the dataset). The result should look like this:
Investigating a relationship
Let’s say we’re curious about the relationship between glucose and hba1c.
What’s going on here? It seems like there are two clusters.
What is driving this clustering? Age? Sex? Race? Diabetes?
Aesthetics
Aesthetics aren’t just for mapping columns to the x- and y-axis
You can also use them to assign color, for instance
ggplot(nhanes) +geom_point(aes(x = glucose, y = hba1c,color = diabetes))
Aesthetics aren’t just for mapping columns to the x- and y-axis
We could have used a shape
ggplot(nhanes) +geom_point(aes(x = glucose, y = hba1c, shape=diabetes ))
Or size
ggplot(nhanes) +geom_point(aes(x = glucose, y = hba1c, size=race ))
This one doesn’t really make sense because we’re mapping a categorical variable to an aesthetic that can take continuous values that imply some ordering
If we set a property outside of the aesthetic, it no longer maps that property to a column.
ggplot(nhanes) +geom_point(aes(x = glucose, y = hba1c ),color ="blue" )
However, we can use this to assign fixed properties to the plot that don’t depend on the data
Exercise: Plot
Can you recreate this plot?
Exercise [together]
What will this do? Why?
ggplot(nhanes) +geom_point(aes(x = glucose, y = hba1c, color ="blue"))
Geoms
ggplot(nhanes) +geom_point(aes(x = glucose, y = hba1c))
ggplot(nhanes) +geom_smooth(aes(x = glucose, y = hba1c))
Both these plots represent the same data, but they use a different geometric representation (“geom”)
e.g. bar chart vs. line chart, etc.
R graph gallery is a great resource to help you design your plot and pick the right geom: r-graph-gallery.com
Different geoms are configured to work with different aesthetics.
e.g. you can set the shape of a point, but you can’t set the “shape” of a line.
On the other hand, you can set the “line type” of a line:
ggplot(nhanes) +geom_smooth(aes(x = glucose, y = hba1c, linetype = diabetes))
Use e.g. ?geom_point to see what aesthetics are expected or allowed
It’s possible to add multiple geoms to the same plot
ggplot(nhanes) +geom_smooth(aes(x = glucose, y = hba1c, color = diabetes)) +geom_point(aes(x = glucose, y = hba1c, color = diabetes))
To assign the same aesthetics to all geoms, pass the aesthetics to the ggplot function directly instead of to each geom individually
ggplot(nhanes, aes(x = glucose, y = hba1c, color = diabetes)) +geom_smooth() +geom_point()
You can also use different mappings in different geoms
Use google or other resources to figure out how to receate this plot in R:
ggplot(nhanes) +
...
What might the name of this geom be? What properties of the plot (aesthetics) are mapped to what columns of the data?
If you accomplish making the plot, can you figure out how to change the colors of the groups?
Facets
Aesthetics are useful for mapping columns to particular properties of a single plot
Use facets to generate multiple plots with shared structure
ggplot(nhanes) +geom_point(aes(x = glucose, y = hba1c)) +facet_wrap(vars(diabetes), nrow =2)
facet_wrap is good for faceting according to unordered categories
facet_grid is better for ordered categories, and can be used with two variables
ggplot(nhanes) +geom_point(aes(x = glucose, y = hba1c)) +facet_grid(rows=vars(diabetes), cols=vars(sex))
Exercise
Use ggplot to investigate the relationship between these measurements and diabetes using any combination of any kinds of plots that you like. Which measurements are most associated with diabetes? Does this vary by race, age, or assigned sex at birth?
For some plots it may be helpful to reformat your data using this code (we’ll learn how to do this on day 4):
# A tibble: 354 x 7
id age sex race diabetes measure value
<chr> <dbl> <chr> <chr> <chr> <chr> <dbl>
1 NHANES-100129 72 male Mexican American Yes glucose 124
2 NHANES-100129 72 male Mexican American Yes hba1c 6.8
3 NHANES-100129 72 male Mexican American Yes bmi 31.5
4 NHANES-100129 72 male Mexican American Yes waist_cm 107.
5 NHANES-100129 72 male Mexican American Yes bp_sys_1 124
6 NHANES-100129 72 male Mexican American Yes cholesterol 170
7 NHANES-100188 80 female White Yes glucose 95
8 NHANES-100188 80 female White Yes hba1c 6.7
9 NHANES-100188 80 female White Yes bmi 26
10 NHANES-100188 80 female White Yes waist_cm 87
# i 344 more rows
