1 Day 1: Setup and First Steps in R
1.1 Learning objectives
By the end of this day you should be able to:
- Install R and RStudio on macOS, Windows, or Linux.
- Distinguish the console (interactive) from a script (saved code).
- Create and assign variables; perform arithmetic.
- Recognize and work with R’s core data types: numeric, integer, character, logical, factor.
- Index vectors by position, by name, and by logical condition.
- Recognize missing values (
NA) and avoid the common pitfalls.
1.2 Lecture
R is a programming language built specifically for statistical computing. You will spend the next few years of your graduate career running R in some form nearly every day. This first hour has a modest goal: R installed, expressions typed and evaluated, and a working sense of what R is doing with the characters you type.
1.2.1 Installing R and RStudio
Two pieces of software, in this order.
R itself is the language and the engine. Download from https://cran.r-project.org/ and run the installer for your operating system. R 4.5+ is current as of writing; any 4.x release is fine for this book.
RStudio is the integrated development environment (IDE). It is the application you actually open every day; it talks to R behind the scenes. Download from https://posit.co/download/rstudio-desktop/ and install.
When you open RStudio for the first time you see four panes:
- Source (top-left): your code, saved in
.Rfiles. - Console (bottom-left): the live R session.
- Environment / History (top-right): the objects you have created, and a log of recent commands.
- Files / Plots / Help / Packages (bottom-right): filesystem, figures you have produced, help pages, installed packages.
You will use the source pane (writing scripts) and the console pane (running them) most of the time.
1.2.2 The console: arithmetic and assignment
Click in the Console pane (bottom-left). You see a prompt (>). Type:
2 + 2
#> [1] 4R evaluated 2 + 2 and printed 4. The [1] means ‘the first element of the result is printed here’, which matters when results have many elements.
The arithmetic operators are what you expect: +, -, *, /, ^ for exponentiation, %% for modulo (remainder).
17 %% 5
#> [1] 2To save a value for later, assign it to a name. The R convention is to use <- (left-arrow):
x <- 17
x
#> [1] 17
x * 2
#> [1] 34Names can use letters, digits, dot, and underscore, but must start with a letter or dot. The convention in modern R is snake_case:
patient_age <- 47
mean_blood_pressure <- 128.51.2.3 Vectors
The fundamental data structure in R is the vector: an ordered collection of values of the same type. Create one with c() (combine):
ages <- c(24, 47, 31, 62, 19, 55)
ages
#> [1] 24 47 31 62 19 55
length(ages)
#> [1] 6Arithmetic on a vector applies element-wise:
ages_in_months <- ages * 12
ages_in_months
#> [1] 288 564 372 744 228 660Many R functions take a vector and return either a vector of the same length or a single summary value:
mean(ages)
#> [1] 39.66667
sd(ages)
#> [1] 17.36856
range(ages)
#> [1] 19 62This vectorized behavior is the single most important idiom in R. You will rarely need to write a loop to apply an operation element by element. R does it for you.
1.2.4 Data types
Every value in R has a type. The five you will meet this week:
typeof(42L) # integer (note the L)
#> [1] "integer"
typeof(3.14) # double-precision floating-point
#> [1] "double"
typeof("hello") # character (string)
#> [1] "character"
typeof(TRUE) # logical (boolean)
#> [1] "logical"
typeof(factor(c("low", "high"))) # factor
#> [1] "integer"A factor is R’s way of representing a categorical variable. Internally it is an integer with a labeling. You will encounter factors immediately when reading data. For now, treat them simply as categorical.
R coerces between types automatically when it can:
TRUE + 1
#> [1] 2 # TRUE is treated as 1
"5" + 1
#> Error: non-numeric argument to binary operator
as.numeric("5") + 1
#> [1] 6Coercion is the source of many beginner bugs. When in doubt, it is worth checking the type directly with typeof() or class().
1.2.5 Indexing vectors
Three ways to extract elements from a vector.
By position (integer index, starting at 1):
ages <- c(24, 47, 31, 62, 19, 55)
ages[1]
#> [1] 24
ages[c(1, 3, 5)]
#> [1] 24 31 19
ages[-1] # all but the first
#> [1] 47 31 62 19 55By logical (a vector of TRUE/FALSE of the same length):
adults <- ages >= 18
adults
#> [1] TRUE TRUE TRUE TRUE TRUE TRUE
ages[ages >= 50]
#> [1] 62 55By name (when the vector has names):
biomarkers <- c(hba1c = 7.2, ldl = 145, sbp = 132)
biomarkers["ldl"]
#> ldl
#> 145Logical indexing is the workhorse of data manipulation in R. It is best read as ‘keep the elements where the condition is TRUE’.
1.2.6 Missing values
Real biomedical data has missing values, and it is worthwhile to become comfortable with them early. R represents a missing value as NA (Not Available):
labs <- c(7.2, NA, 6.8, 8.1, NA)
labs
#> [1] 7.2 NA 6.8 8.1 NA
mean(labs)
#> [1] NA
mean(labs, na.rm = TRUE)
#> [1] 7.366667mean() returns NA by default when any input is NA. Adding na.rm = TRUE removes the missing values before computing. Most R functions follow this convention, and the convention is deliberate. It forces you to decide how to handle missingness rather than silently ignoring it.
Test for missingness with is.na():
is.na(labs)
#> [1] FALSE TRUE FALSE FALSE TRUE
labs[!is.na(labs)]
#> [1] 7.2 6.8 8.1Do not test for NA with ==:
labs == NA
#> [1] NA NA NA NA NANA == NA is itself NA. The result of comparing an unknown quantity to an unknown quantity is, appropriately, unknown.
1.3 Worked example: a small biomedical dataset
You see ten patients in a clinic. Their ages and systolic blood pressures (SBP) are recorded.
patient_id <- c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10)
ages <- c(34, 67, 52, 41, 58, 23, 71, 49, NA, 38)
sbp <- c(122, 148, 135, 128, 142, 118, 156, 130,
139, 124)Sample size, mean age (handling the missing), proportion hypertensive (SBP >= 140):
length(patient_id)
#> [1] 10
mean(ages, na.rm = TRUE)
#> [1] 48.11111
mean(sbp >= 140)
#> [1] 0.3Identify the hypertensive patients and report their IDs and ages:
hypertensive <- sbp >= 140
hypertensive
#> [1] FALSE TRUE FALSE FALSE TRUE FALSE TRUE FALSE FALSE FALSE
patient_id[hypertensive]
#> [1] 2 5 7
ages[hypertensive]
#> [1] 67 58 71In three lines you have done what an analyst might otherwise work through by hand: counted the cohort, computed summary statistics with explicit missing-data handling, and identified the subset of clinical interest. This is the core rhythm of R for biostatistics, and it will recur, in more elaborate form, throughout the week.
1.4 Homework
Each problem is solvable with what was covered in the lecture. Attempt each one in your R session before checking the solution.
Reproduce. In your R session, type all of the examples from the ‘Worked example’ section. Confirm you get the same output.
Five small expressions. Write R expressions that compute each of these:
- The cube root of 27.
- The remainder when 1000 is divided by 7.
- The mean of the integers 1 through 100.
- The number of values in
c(2, 4, 6, 8, 10)that are greater than 5. - The maximum minus the minimum of
c(11, 4, 19, 7, 2).
Inspect three suspicious expressions. For each, run it in R and explain what happens. If it errors, explain why; if it runs, explain whether the result is what a careful R user would write.
mean(c(1, 2, 3, "4"))c(1, 2, 3) + c(10, 20)x = 5; x ** 2
Index by group. Given
ages <- c(8, 22, 47, 65, 13, 51, 73, 29, 17, 55), compute the mean age in three groups: children (under 18), working-age adults (18-64 inclusive), seniors (65+). Use logical indexing.Missing values. Given
labs <- c(7.2, NA, 6.8, 8.1, NA, 5.9, 7.4, NA):- How many values are missing?
- What is the mean of the non-missing values?
- Replace each
NAwith the mean of the non-missing values, producing a vector of length 8 with noNAs.
1.5 Solutions
Problem 1. Each line in the worked example should produce the output shown. If yours differs, common causes are: typo (check parentheses and commas), wrong type (e.g., quotes around numbers), or a left-over object from earlier in the session. Restart R if the session is in a strange state.
Problem 2.
# (a)
27 ^ (1/3)
#> [1] 3
# (b)
1000 %% 7
#> [1] 6
# (c)
mean(1:100)
#> [1] 50.5
# (d)
sum(c(2, 4, 6, 8, 10) > 5)
#> [1] 3
# (e)
max(c(11, 4, 19, 7, 2)) - min(c(11, 4, 19, 7, 2))
#> [1] 17
# (or, more cleanly:)
diff(range(c(11, 4, 19, 7, 2)))
#> [1] 17Problem 3.
mean(c(1, 2, 3, "4")). The"4"(in quotes) is a character;c()coerces the whole vector to character because mixed types must collapse to one type.mean()on a character vector errors. Fix:
mean(c(1, 2, 3, 4))
#> [1] 2.5c(1, 2, 3) + c(10, 20). R recycles the shorter vector, but lengths that do not divide evenly produce a warning. Fix to matching lengths:
c(1, 2, 3) + c(10, 20, 30)
#> [1] 11 22 33x = 5; x ** 2. The=is allowed for assignment, but the convention in R is<-. R’s parser silently rewrites**to^(see?Arithmetic), so this expression returns 25. The expression is not strictly broken, but**is not idiomatic in R and relies on a parser quirk that is easy to forget. We prefer^explicitly:
x <- 5
x ^ 2
#> [1] 25Problem 4.
ages <- c(8, 22, 47, 65, 13, 51, 73, 29, 17, 55)
mean(ages[ages < 18]) # children
#> [1] 12.66667
mean(ages[ages >= 18 & ages < 65]) # working-age
#> [1] 40.8
mean(ages[ages >= 65]) # seniors
#> [1] 69Problem 5.
labs <- c(7.2, NA, 6.8, 8.1, NA, 5.9, 7.4, NA)
# (a)
sum(is.na(labs))
#> [1] 3
# (b)
mean(labs, na.rm = TRUE)
#> [1] 7.08
# (c)
labs[is.na(labs)] <- mean(labs, na.rm = TRUE)
labs
#> [1] 7.20 7.08 6.80 8.10 7.08 5.90 7.40 7.08A caution is in order here. Imputing missing values with the mean is sometimes acceptable for exploratory work, but it can bias inferences in a real analysis. Day 4 introduces statistical tests. The Practicum and Applied Methods volumes treat missing-data handling at the depth publishable work requires.
1.6 What’s next
Day 2 covers reading real data into R and manipulating it with the dplyr verbs. Bring your laptop with R and RStudio installed and tested.