Describing Lab

Today’s job

Describing Data, 75 points, due this week

• Tuesday you watched me build three figures on my variables
• Today you build three figures on yours, with me in the room
• Everything you make today lands in your White Paper’s Results section
• Work in your project folder, in a .qmd file, not in the console
• Get stuck out loud. That is what the room is for.

Where you should be starting

• Your project opens in VS Code and your .qmd renders without error
• You have a tidy analysis table from the wrangling assignment
• You know which one variable your paper is really about
• You know which column splits it into groups
• If any of those is missing, say so now and we fix it first

Loading and looking

library(tidyverse)
library(v2v)

analysis <- readRDS("data/analysis.rds")

glimpse(analysis)

glimpse() prints every column with its type and its first few values. Before you plot anything, look at what you actually have.

A summary table first

analysis %>%
  group_by(is_gaming) %>%
  summarise(
    n      = n(),
    mean   = mean(message_length, na.rm = TRUE),
    median = median(message_length, na.rm = TRUE),
    sd     = sd(message_length, na.rm = TRUE),
    .groups = "drop"
  )

Swap is_gaming for your grouping column and message_length for your variable. Read the mean against the median before you draw anything.

Figure one: the shape

ggplot(analysis, aes(x = message_length)) +
  geom_histogram(binwidth = 5) +
  labs(x = "Message length (characters)", y = "Messages") +
  v2v::theme_v2v()

One variable, one histogram. Change binwidth and watch the shape change: that is why the number has to be reported, not hidden.

Reading your own histogram

• Where is the tall stack? That is where the typical case lives.
• Is there a long thin tail? If so, in which direction?
• Right-skewed means most values are small and a few are very large
• A skewed variable is why your mean and median disagreed
• If you capped the axis with pmin(), say so in the axis label

Figure two: counts by category

analysis %>%
  count(game, name = "messages") %>%
  slice_max(messages, n = 8) %>%
  ggplot(aes(x = reorder(game, messages), y = messages)) +
  geom_col(fill = "#2f7d8a") +
  coord_flip() +
  v2v::theme_v2v()

Count, keep the top eight, sort the bars, then flip the chart so long category names are readable. Unsorted bars are much harder to compare.

Figure three: groups side by side

analysis %>%
  filter(!is.na(is_gaming)) %>%
  ggplot(aes(x = message_length, fill = is_gaming)) +
  geom_histogram(binwidth = 5, position = "identity", alpha = 0.55) +
  v2v::scale_fill_v2v() +
  v2v::theme_v2v()

position = "identity" with alpha overlays the two groups so you can compare shapes. Stacked bars would hide exactly the comparison you want.

Labels are not optional

last_plot() +
  labs(
    title = "Message length by channel type",
    x = "Message length (characters, capped at 120)",
    y = "Messages",
    fill = "Gaming channel"
  )

Default labels are column names, which mean nothing to a reader. Every axis gets units. Every cap or transformation gets named in the label.

Colorblind-safe by default

• Roughly one man in twelve has a color-vision deficiency
• Red against green is the notorious pair, and it is not the only one
• Use scale_colour_v2v() and scale_fill_v2v(), which are chosen to stay distinct
• Do not let color carry the message alone: add a linetype, or split into panels
• Test it: if the figure fails in grayscale, it fails for some of your readers

Alt text, written not skipped

• A screen-reader user cannot see the figure. Alt text is what they get instead.
• A figure with no alt text is, to them, simply missing
• Alt text is not the caption repeated
• Say what kind of chart, what is on each axis, and what it shows
• In Quarto it goes in the figure’s fig-alt attribute

Common errors and what they mean

• object 'analysis' not found: you never ran the line that loads it
• could not find function "ggplot": library(tidyverse) has not run
• Blank grey panel: your aes() names a column that does not exist
• Error in +: a + is missing at the end of a line, or one is trailing at the end
• Everything on one bar: your x variable is a character, not a number

The package ships the ones specific to this data: run ?v2v::common_errors for the year-50,888 timestamp, the join that matches nothing, and the surprisingly low kappa.

Saving figures for the paper

ggsave(
  "figures/fig-msglen.png",
  width = 7, height = 4.5, dpi = 300
)

ggsave() writes the last plot to a file at a size and resolution you control. Fixed dimensions mean your figure looks the same everywhere, not squashed.

Checkpoint

You should now have, saved in your project:

• A grouped summary table with n, mean, median, and sd
• Three figures, all using theme_v2v(), all fully labeled
• Alt text drafted for each one
• One sentence under each figure saying what it shows
• A .qmd that renders start to finish without error

Before next time

• Describing Data [R] is due this week, 75 points
• Read Chapter 13, “Making the call,” before Tuesday
• Tuesday we answer the question your figures could only frame: is the gap real?
• Bring the summary table you built today. We test the difference in it.
• No new math to learn in advance. Come rested, come curious.