Learn

The tidyverse is a collection of packages by Hadley Wickham that defines a consistent grammar for data manipulation. The packages: dplyr (transform), tidyr (reshape), ggplot2 (plot), readr (read), purrr (functional), stringr (strings), forcats (factors), lubridate (dates).

The pipe |> (or %>% from magrittr): x |> f() = f(x). Chain transformations:

library(dplyr)

library(dplyr)
result <- mtcars |>
    filter(cyl == 6) |>
    mutate(efficiency = mpg / hp) |>
    group_by(gear) |>
    summarise(avg_eff = mean(efficiency)) |>
    arrange(desc(avg_eff))

Reads top-to-bottom: take mtcars, keep 6-cylinder rows, add an efficiency column, group by gear, summarise mean efficiency, sort descending.

Core dplyr verbs:

• filter(cond) — keep rows where cond is TRUE
• select(cols) — pick columns
• mutate(new = expr) — add/modify columns
• arrange(col) — sort rows; desc(col) reverses
• group_by(col) + summarise(stat = fn(col)) — aggregate
• slice(n) — pick rows by index
• distinct() — drop duplicates
• rename(new = old) — rename column
• count(col) — group_by + summarise(n())
• *_join(other) — inner_join, left_join, right_join, full_join, anti_join

mutate with conditionals:

df |> mutate(
    grade = case_when(
        score >= 90 ~ "A",
        score >= 80 ~ "B",
        score >= 70 ~ "C",
        TRUE ~ "F"
    )
)

Pivot wider/longer (tidyr):

library(tidyr)

# Long format -> wide format
long |> pivot_wider(names_from = quarter, values_from = sales)

# Wide format -> long format
wide |> pivot_longer(cols = q1:q4, names_to = "quarter", values_to = "sales")

Long format is usually what you want for analysis. Wide format is what spreadsheets give you.

ggplot2 — the grammar of graphics:

library(ggplot2)

ggplot(mtcars, aes(x = wt, y = mpg, color = factor(cyl))) +
    geom_point() +
    geom_smooth(method = "lm") +
    labs(title = "MPG vs Weight by Cylinders", color = "Cyl")

Layer-based: data + aesthetic mappings + geometric layers + scale/theme.

Why tidyverse?

• Consistent: every verb is data-first, returns a data frame, takes column names unquoted
• Pipeable: |> chains naturally
• Documented: each package has comprehensive vignettes
• Active: still evolving with tidymodels (ML), targets (workflows), arrow (Parquet)

When to skip:

• Tiny scripts where base R is shorter
• Performance-critical code (use data.table)
• Libraries that need minimal dependencies

For analysis work, tidyverse is the dominant idiom — most R job postings list dplyr+ggplot2.