Tidyverse Tips / Refresher

Reading CSV files: read_csv() vs. read.csv()

Code 
# Base R
data_base <- read.csv("my_data.csv")

# Tidyverse (readr)
library(tidyverse)
data_tidy <- read_csv("my_data.csv")

Why prefer read_csv()?

  • Automatically parses column types (numeric, date, etc.) and shows a progress bar.
  • Returns a tibble, which prints more cleanly (truncated rows/columns).
  • Faster for large files.

Tip: To store data locally, use write_csv(data, "out.csv").

Inspecting Your Data Immediately

Code 
head(data_tidy)     # first 6 rows
tail(data_tidy)     # last 6 rows
str(data_tidy)      # structure: types & sample values
glimpse(data_tidy)  # tidyverse-friendly structure
dim(data_tidy)      # rows, columns
names(data_tidy)    # column names

Why inspect early?

  1. Check that columns imported with correct (or at least expected) names and types
  2. Spot missing values or parsing problems
  3. Get a sense of dataset size and structure

Dealing with Spaces or Special Characters in Names

Sometimes column names contain spaces or punctuation. You can’t refer to them directly without backticks.

Code 
# Suppose “Number of deaths” was imported:
data$`Number of deaths`

# Better: rename immediately
data <- data %>%
  rename(NumberOfDeaths = `Number of deaths`)

Tip: Use janitor’s clean_names() to automatically convert all names to snake_case:

Code 
library(janitor)
data <- data %>% clean_names()
# “Number of Deaths” → number_of_deaths

Renaming Columns

Code 
data <- data %>%
  rename(
    deaths_total = `Number of deaths`,
    country_code = CountryCode
  )

Rename multiple columns in one call using new_name = old_name syntax.

Merging (Joining) Two Datasets

Why We Merge

  • Merge (join) means adding columns by matching rows.
  • Requires a common identifier (key) that uniquely matches rows across datasets.

Join Functions

Code 
# Simple one-key join
merged <- main_data %>%
  left_join(data_to_add, by = "country_year_id")

# Two-key join
merged <- main_data %>%
  left_join(data_to_add, by = c("country", "year"))

# Other types:
# inner_join(): only keep rows present in both
# right_join(): keep all from data_to_add

Tip: Before joining, ensure keys have the same type and values:

Code 
unique(main_data$country)
unique(data_to_add$country)

Tip: If one dataset uses “DEU” and the other “Germany,” recode or create a lookup table before joining.

Appending (Binding) Rows

This is how we would stack observations:

Code 
# Recommended
total <- bind_rows(data_for_germany, 
                   data_for_france)

# Base R equivalent:
total2 <- rbind(data_for_germany, data_for_france)

Note: bind_rows() will fill in missing columns with NA if one data frame has extra columns.

Common Operations

  1. Chaining with the pipe %>%
Code 
result <- data %>%
     filter(year >= 2000) %>%
     select(country, year, deaths_total) %>%
     arrange(desc(deaths_total))
  1. Use mutate() to create or transform columns
Code 
data <- data %>%
     mutate(
       deaths_per_100k = deaths_total / population * 100000,
       log_deaths = log(deaths_total)
     )
  1. Quick summaries with group_by() + summarise()
Code 
summary <- data %>%
     group_by(country) %>%
     summarise(
       total_deaths = sum(deaths_total, na.rm = TRUE),
       avg_deaths = mean(deaths_total, na.rm = TRUE)
     )
  1. Check for duplicates (especially before joining or binding)
Code 
data %>%
     add_count(country, year) %>%
     filter(n > 1)
  1. Convert to factors or dates cleanly
Code 
data <- data %>%
     mutate(
       country = as_factor(country),
       date = lubridate::ymd(date_string)
     )
  1. Use glimpse() for a prettier, horizontally oriented overview.

Useful reference

See Grant McDermott’s excellent slides