Introduction to R

Pakistan District MPI, 2019-20

D4D | Data for Development

Today’s Data

Pakistan District-level MPI, 2019-20 — 127 districts

Three core variables:

• mpi — overall poverty score (higher = poorer)
• h_incidence — H: % of population that is poor
• a_intensity — A: average deprivation among the poor

Relationship: MPI = H × A / 100

Step 1: Load Tools

library(tidyverse)

One package. Everything we need today lives inside it.

Step 2: Import Data

mpi <-  here::here("data", "pakistan_mpi_district_2019_20.csv")  # check file path)

read_csv() reads the file and stores it as mpi.

Step 3: First Look at the Data

dim(mpi)        # rows x columns
str(mpi)        # column names and types
head(mpi)       # first 6 rows
glimpse(mpi)    # compact structure view

Run these four commands on every new dataset.

Step 4: Further Inspection

tail(mpi)       # last 6 rows
summary(mpi)    # min / mean / max for numeric columns

Step 5: Standardise Column Names

mpi <- mpi |> janitor::clean_names()
names(mpi)

Good practice for every imported dataset, regardless of source.

Step 6: Focus on the Three Core Variables

mpi |>
  select(district, mpi, h_incidence, a_intensity) |>
  head(10)

select() narrows the data to the columns we need.

Step 7: Rank Districts by MPI

mpi |>
  select(district, province, mpi) |>
  arrange(desc(mpi)) |>
  head(10)

arrange(desc()) sorts from highest to lowest.

Step 8: Filter by Province

mpi |>
  filter(province == "Balochistan") |>
  select(district, mpi, h_incidence, a_intensity) |>
  arrange(desc(mpi))

filter() keeps only rows meeting a condition.

Step 9: Province-Level Averages

mpi |>
  summarise(
    avg_mpi = mean(mpi),
    avg_h   = mean(h_incidence),
    avg_a   = mean(a_intensity),
    .by = province
  ) |>
  arrange(desc(avg_mpi))

.by = groups without a separate group_by() / ungroup() pair.

Step 10: Build a Plot — The Canvas

ggplot(mpi, aes(x = h_incidence, y = a_intensity))

No geometry mapped yet — this produces an empty canvas.

Step 10: Build a Plot — Add Points

ggplot(mpi, aes(x = h_incidence, y = a_intensity)) +
  geom_point()

Each point now represents one district.

Step 10: Build a Plot — Style and Label

ggplot(mpi, aes(x = h_incidence, y = a_intensity, color = province)) +
  geom_point(size = 2) +
  theme_minimal() +
  labs(
    title    = "District-Level MPI Components, 2019-20",
    subtitle = "Each point represents one district",
    x        = "H: Incidence of poverty (%)",
    y        = "A: Intensity of deprivation",
    color    = "Province"
  )

Step 11: Bar Chart of the 10 Poorest Districts

top10 <- mpi |>
  arrange(desc(mpi)) |>
  head(10)

ggplot(top10, aes(x = reorder(district, mpi), y = mpi, fill = province)) +
  geom_col() +
  coord_flip() +
  theme_minimal() +
  labs(
    title = "10 Poorest Districts by MPI Score (2019-20)",
    x = NULL,
    y = "MPI Score"
  )

Step 12: Export Outputs

province_summary <- mpi |>
  summarise(
    avg_mpi = mean(mpi),
    avg_h   = mean(h_incidence),
    avg_a   = mean(a_intensity),
    .by = province
  )

write_csv(province_summary, "province_mpi_summary.csv")
ggsave("top10_poorest_districts.png", width = 8, height = 5)

Summary of Commands Covered

Task	Commands
Import	read_csv()
Inspect	dim(), str(), head(), tail(), glimpse(), summary()
Clean	clean_names()
Transform	select(), filter(), arrange(), summarise(), .by
Visualize	ggplot(), geom_point(), geom_col(), theme_minimal()
Export	write_csv(), ggsave()

Next

Same workflow, any column in the dataset.

Try it yourself: repeat Steps 6–9 using pop_share or n_poor_thousands.