Subsetting in R

Author

Affiliation

Julianne G. Clina

University of Kansas Medical Center

Published

August 26, 2026

Learning Objectives

By the end of this workshop, you should be able to:

Explain the difference between [, [[, and $, and know when to use each
Subset a data frame by rows and columns using data[rows, columns]
Explain the difference between simplifying and preserving subsetting
Build logical conditions with ==, >=, <, &, |, and ! to select rows
Combine multiple conditions (including with parentheses) to answer a realistic research question
Translate a plain-English data request into R subsetting code

The core question to ask yourself throughout: what information do I want, and where is it stored?

##Creating Example Dataset for practice!
library(tidyverse)


kids <- tibble(
  id = sprintf("K%02d", 1:15),
  age = c(
    8,
    9,
    10,
    8,
    11,
    12,
    9,
    10,
    11,
    8,
    12,
    9,
    10,
    11,
    12
  ),
  sex = c(
    "F",
    "M",
    "F",
    "M",
    "F",
    "M",
    "F",
    "F",
    "M",
    "M",
    "F",
    "M",
    "F",
    "M",
    "F"
  ),
  bmi_z = c(
    -0.2,
    0.8,
    1.4,
    -0.5,
    2.1,
    0.3,
    1.1,
    -0.8,
    0.6,
    1.8,
    0.2,
    1.3,
    -0.1,
    2.4,
    0.9
  ),
  mvpa_min = c(
    62,
    41,
    28,
    75,
    35,
    54,
    92,
    67,
    46,
    22,
    81,
    38,
    70,
    31,
    55
  ),
  sedentary_min = c(
    510,
    560,
    610,
    480,
    590,
    525,
    450,
    500,
    545,
    625,
    470,
    580,
    495,
    600,
    530
  ),
  steps = c(
    9200,
    7100,
    5400,
    10400,
    6300,
    8100,
    12000,
    9800,
    7600,
    4900,
    11000,
    6800,
    10100,
    5700,
    8300
  ),
  sbp = c(
    102,
    108,
    115,
    98,
    121,
    110,
    105,
    100,
    112,
    118,
    107,
    114,
    103,
    124,
    109
  ),
  dbp = c(
    64,
    68,
    72,
    61,
    76,
    70,
    66,
    62,
    69,
    74,
    67,
    71,
    65,
    78,
    68
  ),
  resting_hr = c(
    82,
    78,
    88,
    76,
    91,
    80,
    75,
    79,
    84,
    94,
    77,
    86,
    74,
    90,
    81
  ),
  vo2peak = c(
    42.1,
    38.4,
    34.2,
    45.6,
    31.8,
    40.2,
    47.1,
    43.5,
    37.9,
    29.4,
    44.8,
    35.6,
    46.2,
    32.5,
    39.7
  )
)

kids$parent_info <- list(
  list(mother = "Sarah", father = "James", contact = "email"),
  list(mother = "Maria", father = "Carlos", contact = "phone"),
  list(mother = "Jennifer", father = "David", contact = "email"),
  list(mother = "Amanda", father = "Brian", contact = "phone"),
  list(mother = "Rachel", father = "Michael", contact = "email"),
  list(mother = "Lisa", father = "Robert", contact = "email"),
  list(mother = "Emily", father = "Daniel", contact = "phone"),
  list(mother = "Katie", father = "Andrew", contact = "email"),
  list(mother = "Nicole", father = "Matthew", contact = "phone"),
  list(mother = "Jessica", father = "Christopher", contact = "email"),
  list(mother = "Laura", father = "Steven", contact = "phone"),
  list(mother = "Michelle", father = "Kevin", contact = "email"),
  list(mother = "Elizabeth", father = "Mark", contact = "email"),
  list(mother = "Heather", father = "Jason", contact = "phone"),
  list(mother = "Rebecca", father = "Thomas", contact = "email")
)

kids$mvpa_min[c(5, 12)] <- NA
kids$sbp[9] <- NA
kids$bmi_z[13] <- NA

kids

Activity 1: `[`, `[[`, and `$`

Why three operators?

R gives you three tools for pulling things out of an object, and they are not interchangeable:

Operator	What it returns	Typical use
`[`	A subset of the same type/structure (can hold 0, 1, or many elements)	“Give me a smaller version of this object”
`[[`	A single element, unwrapped from its container	“Give me the contents of this one slot”
`$`	A single named element (shorthand for `[[` with a name)	“Give me the column/element called ___”

A helpful mental picture from Advanced R (Wickham): if a list is a train, and each element is a train car, then:

x[4:6] gives you a smaller train made of cars 4 through 6 (still a train — a list)
x[[5]] opens car 5 and hands you what’s inside it (not a train anymore)

This is why [ on a list always returns a list (even a list of length 1), while [[ and $ reach inside and hand you the actual object stored there.

Our parent_info column is a list-column — each element of kids$parent_info is itself a list. It’s a good place to see [ vs [[ vs $ in action.

Step 1: Look at the list-column

kids$parent_info

Step 2: Extract one child’s parent information

Suppose we want the parent information for child 3.

# Use [ ] to extract the third element of parent_info

# Now use [[ ]] to extract the same element

# Compare the classes of the two results

What’s different? [ keeps the “train car” wrapper — you get a list containing one list. [[ unwraps it and gives you the list itself. This is the simplify vs. preserve distinction: [[ simplifies (drops a layer of structure), [ preserves it.

Step 3: Extract information within the list

Now that we have the parent information for child 3, we want to extract just the mother’s name. Because parent_info[[3]] is a named list, we can use $ to grab a named piece of it.

# Extract the mother's name for child 3

The key idea: these operators chain together to walk through nested objects — first [[ to get into the list-column, then $ to get the named element inside.

Activity 2: Selecting Rows and Columns

Data frames behave like a cross between a list (each column is an element) and a matrix (rows and columns). When subsetting with two indices, the pattern is:

data[rows, columns]

The first position (before the comma) says which rows
The second position (after the comma) says which columns
Leaving a position blank means “give me all of them”

# All columns for child 5
kids[5, ]

# The mvpa_min column, for all children
kids[, "mvpa_min"]

Your turn

Imagine you are preparing a small table for a research meeting. You want information for children 2, 5, 8, and 12 — but only the columns id, age, sex, mvpa_min, and vo2peak.

# Select children 2, 5, 8, and 12
# and keep only id, age, sex, mvpa_min, and vo2peak

Simplify vs. preserve: one more thing to notice

Compare these two approaches to grabbing a single column:

kids[, "mvpa_min"] # matrix-style: simplifies
kids["mvpa_min"] # list-style: preserves

They select the same values, but return different types of objects:

class(kids[, "mvpa_min"])
class(kids["mvpa_min"])

kids[, "mvpa_min"] simplifies down to a plain vector, because you asked for it “matrix style” (row, column). kids["mvpa_min"] treats kids like a list and hands you back a smaller data frame with one column, preserving the original structure. This distinction is a common source of bugs — a function that works fine when you select multiple columns can break unexpectedly when only one column is selected, because the object type silently changes. (Tibbles were designed to avoid this: kids[, "mvpa_min"] on a tibble still returns a tibble, not a bare vector — try it and compare to a base data.frame.)

Activity 3: Logical Subsetting

So far we’ve selected rows by their position (row 5, rows 2/5/8/12). But in research data, we usually want to select rows by whether they meet a condition — “give me children who did X” rather than “give me row 7.”

This is where logical subsetting comes in. A logical condition like:

kids$mvpa_min >= 60

produces a vector of TRUE, FALSE, and possibly NA — one value per row. When you put that logical vector inside [, R keeps only the rows where the value is TRUE, and returns NA for rows where the condition itself was NA (this is different from using which(), which silently drops NAs instead — worth knowing when you’re not sure why your row counts don’t match).

Your turn

Find the children who get at least 60 minutes of MVPA. Return only id, age, mvpa_min, and vo2peak.

# Select children with at least 60 minutes of MVPA
# Keep only id, age, mvpa_min, and vo2peak

Try another one

Now find just the IDs of children who get less than 30 minutes of MVPA.

# Select the IDs of children with MVPA < 30

Notice this time you’re subsetting a vector (kids$id), not the whole data frame — same logical-subsetting idea, just applied outside the data[rows, columns] pattern.

Activity 4: Combining Conditions

Multiple logical conditions can be combined with:

Operator	Meaning
`&`	AND (both conditions must be TRUE)
`\|`	OR (at least one condition must be TRUE)
`!`	NOT (flips TRUE/FALSE)

Always use the vectorized & and | here (not && / ||, which only compare a single value each and are meant for if statements).

kids$mvpa_min >= 60 &
  kids$bmi_z < 1

Your turn

Identify children who have either low physical activity (MVPA < 30 minutes) or high BMI z-score (≥ 2). Return id, age, bmi_z, mvpa_min.

# Select children with either:
#   MVPA < 30
#   OR BMI z-score >= 2
#
# Keep id, age, bmi_z, and mvpa_min

Add another condition

Now restrict to girls who meet either of those criteria:

Female AND (low MVPA OR high BMI z-score)

# Select girls who either:
#   have MVPA < 30
#   OR have BMI z-score >= 2
#
# Keep id, age, sex, bmi_z, and mvpa_min

Why the parentheses matter: without them, R would evaluate & before | in a way that changes the meaning of the statement (operator precedence). The parentheses force the OR condition to be evaluated as a single unit before it’s combined with the AND. When in doubt, use parentheses to make your intent explicit — even when they’re not strictly required, they make code easier for you (and others) to read later.

Activity 5: Putting It All Together

Your PI sends you this request:

“Can you give me the ID, age, sex, MVPA, BMI z-score, and systolic blood pressure for girls who either get less than 30 minutes of MVPA or have a BMI z-score of 2 or greater? Please exclude anyone with missing systolic blood pressure.”

Before writing any code, translate the request into pieces.

Which rows do we want?

Girls
AND either: MVPA < 30, OR BMI z-score ≥ 2
AND systolic BP is not missing

Which columns do we want?

id, age, sex, mvpa_min, bmi_z, sbp

Your turn

# Your code here

Hint: is.na() returns TRUE where a value is missing. How could you flip that with ! to keep only the rows that are not missing?

Key Takeaways

What object am I working with? (kids, kids$parent_info, kids$id, …)
What do I want to extract?
- [ to subset (keeps the same structure/type — a smaller train)
- [[ to extract a single element, unwrapped (open one car)
- $ to extract a single named element (shorthand for [[)
Which rows do I want? Position (1:5, c(2, 5, 8)) or a logical condition (kids$mvpa_min >= 60)
Which columns do I want? By name: c("id", "age", "mvpa_min")

Learning Objectives

Activity 1: [, [[, and $

Why three operators?

Step 1: Look at the list-column

Step 2: Extract one child’s parent information

Step 3: Extract information within the list

Activity 2: Selecting Rows and Columns

Your turn

Simplify vs. preserve: one more thing to notice

Activity 3: Logical Subsetting

Your turn

Try another one

Activity 4: Combining Conditions

Your turn

Add another condition

Activity 5: Putting It All Together

Your turn

Key Takeaways

Activity 1: `[`, `[[`, and `$`