Homework 3

ENVS-193DS

Part 1. Setup tasks

GitHub repository: https://rebeccalmartinez.github.io/ENVS-193DS_homework-03/

# Load packages for data, plots, effects, and animation
library(tidyverse)   # Data wrangling and ggplot2 visualization
library(here)      # Helps keep file paths consistent
library(flextable) # Creates tables
library(janitor)  # Cleans column names 
library(ggtext)    # Styled/markdown text in ggplots
library(ggfx)         # For visual effects for ggplot
library(gganimate)    # Animation for ggplots
library(showtext)     # Render text and emojis in plots better

# Turn on showtext for smoother font/emoji rendering in plots
showtext_auto()

# Optional: Use this only if emojis don't show up correctly in saved plots
font_add("emoji", "seguiemj.ttf")
service_data <- read_csv(here("data", "service_data.csv"))

Part 2. Problems

Problem 1. Personal data

a. Data summarizing

I could count the number of service calls where help was or wasn’t needed, grouped by whether the light was working or broken. Comparing these groups is informative because a broken light might influence whether guests mistakenly request help.

b. Visualizations

Code for cleaning data:

# Clean and standardize help and working columns

clean_service_data <- service_data |>  # using service_data set
  mutate(
    # Convert values to uppercase to avoid issues with lowercase na
    help = toupper(help),
    working = toupper(working),
    # Convert the uppercase text to logical TRUE/FALSE values
     help = as.logical(help), 
    working = as.logical(working))

# Create a stacked bar chart to compare how often help was or wasn't needed
# across two conditions: working lights vs broken lights
ggplot(clean_service_data, aes(x = working, fill = help)) +
  
  # Use stacked bars with slimmer width and black borders for clarity
  geom_bar(position = "stack", width = 0.6, color = "black") +
  # Add counts inside each bar segment
  geom_text(
  # Count the number of rows per group
    stat = "count",
  # Display the count as text
    aes(label = after_stat(count)),
  # Position in each stacked segment
    position = position_stack(vjust = 0.5),
  # text color and size 
   color = "black",
    size = 4
  ) +
  # Assign color to and titles to stacked bars
  scale_fill_manual(
    values = c("TRUE" = "forestgreen", "FALSE" = "red"),
    labels = c("Help Not Needed", "Help Needed")
  ) +
  
  # Replace raw logical values with clearer labels for x-axis categories
  scale_x_discrete(labels = c("TRUE" = "Working", "FALSE" = "Broken")) +
  
  # Force y-axis to start at 0 
  scale_y_continuous(limits = c(0, 50), expand = c(0, 0)) +
  
  # Set axis labels and title
  labs(
    x = "Light Condition",                   
    y = "Number of Service Calls",
    title = "Service Calls by Light Condition and Help Status"
  ) +
  
  # Apply a minimal, clean theme and customize layout
  theme_classic() +
  theme(
  # Center and bold the title
    plot.title = element_text(face = "bold", hjust = 0.5, size = 16), 
  # Bold axis labels
    axis.title = element_text(face = "bold"),                          
  # Move legend to top-right
    legend.position = c(0.95, 0.95),                                   
  # Align legend corner
    legend.justification = c("right", "top"),                          
  # Remove title
    legend.title = element_blank()                                     
  )

c. Caption

Figure 1. False service calls were more common when the light was broken. Stacked bars show the number of service calls under each light condition (Working vs. Broken), split by whether help was actually needed (green) or not (red). Out of 49 broken service lights, 45 did not require help and only 4 did. Out of 43 working service lights, 32 did not require help and 11 did. An additional 16 cases had unknown light status (and are not shown in the figure (gray).

d. Table presentation

#
# Create a summary table that counts how often help was or wasn’t needed,
# grouped by the condition of the service light (working, broken, or unknown)

help_summary <- clean_service_data |>
  mutate(
    # Turn the 'working' column into clear labels for light condition
    light_status = case_when(
      working == TRUE ~ "Working",
      working == FALSE ~ "Broken",
      is.na(working) ~ "Unknown"
    ),
    
    # Turn the 'help' column into clear labels for help status
    help_status = case_when(
      help == TRUE ~ "Help Needed",
      help == FALSE ~ "No Help Needed",
      is.na(help) ~ "Unknown Help"
    )
  ) |>
  
  # Count how many service calls fall into each combination of light and help status
  count(light_status, help_status) |>
  
  # Rearrange the data so each help type becomes its own column
  pivot_wider(
    names_from = help_status,
    values_from = n,
    values_fill = 0  # if a group doesn’t exist, fill it with 0
  )

# Make sure all expected help categories are included (even if not present in the data)
for (col in c("Help Needed", "No Help Needed", "Unknown Help")) {
  if (!col %in% names(help_summary)) {
    help_summary[[col]] <- 0
  }
}

# Order the rows so "Working" comes first and "Unknown" is last,
# then select and reorder columns for readability
help_summary <- help_summary |>
  mutate(light_status = factor(light_status, levels = c("Working", "Broken", "Unknown"))) |>
  arrange(light_status) |>
  select(light_status, `Help Needed`, `No Help Needed`, `Unknown Help`) |>
  
  # Add a new column that totals the calls for each light condition
  mutate(`Total Calls` = `Help Needed` + `No Help Needed` + `Unknown Help`)

# Format the summary table for display using flextable
help_summary |>
  flextable() |>
  set_header_labels(light_status = "Light Condition") |>
  autofit() |>
  set_caption("Table 1. Service calls by light condition and whether help was needed.")

Light Condition	Help Needed	No Help Needed	Unknown Help	Total Calls
Working	11	32	0	43
Broken	4	45	0	49
Unknown	0	0	16	16

Problem 2. Affective visualization

a. Describe in words what an affective visualization could look like for your personal data

I could draw a light board or circuit board that shows which service lights were working and which were not. Each light would be a circle or dot, with different colors based on their status. I would add a symbol like an exclamation point for cases where help was needed, and a question mark for unknowns. It would look more like the data it represents.

b. Create a sketch (on paper) of your idea.

My affective visualization sketch:

c. Make a draft of your visualization

# Clean and transform data to prepare for lightboard
clean_lights <- service_data |>
  mutate(
    # Convert values to logical (TRUE/FALSE/NA)
    help = as.character(help),
    working = as.character(working),
    help = ifelse(help %in% c("T", "TRUE"), TRUE,
           ifelse(help %in% c("F", "FALSE"), FALSE, NA)),
    working = ifelse(working %in% c("T", "TRUE"), TRUE,
              ifelse(working %in% c("F", "FALSE"), FALSE, NA)),

    # Define status labels for light states
    status = case_when(
      is.na(working)     ~ "Unknown",
      working            ~ "Working",
      !working           ~ "Broken"
    ),

    # Assign color to each light status
    color = case_when(
      status == "Working"  ~ "yellow",
      status == "Broken"   ~ "white"
    ),

    # Add emoji symbols to represent help requests or unknowns
    symbol = case_when(
      is.na(working)       ~ "❓",
      help == TRUE         ~ "🛠️",
      TRUE                 ~ ""
    )
  ) |>
  arrange(status)  # Arrange so similar statuses appear together on the board

# Add position and ID for each light on the grid
glow_lights <- clean_lights |>
  mutate(
    light_id = row_number(),
    x = (light_id - 1) %% 9,              # 9 lights per row
    y = -((light_id - 1) %/% 9)           # create rows top to bottom
  )

# Create "on" version (all lights visible)
lights_on <- glow_lights |>
  mutate(blink_state = "on")

# Create "off" version (only hide working lights)
lights_off <- glow_lights |>
  mutate(
    blink_state = "off",
    status = ifelse(status == "Working", NA, status)
  )

# Combine both states for animation
blinking_glow_lights <- bind_rows(lights_on, lights_off)

# Create the blinking lightboard plot
blink_plot <- ggplot(blinking_glow_lights, aes(x, y)) +

  # Add glow effect to working (yellow) lights
  with_outer_glow(
    geom_point(
      data = subset(blinking_glow_lights, status == "Working"),
      aes(fill = status, group = light_id),
      shape = 21,
      size = 6,
      color = "gold2",
      stroke = 0.5,
      na.rm = TRUE
    ),
    colour = "gold3",
    sigma = 5,
    expand = 2
  ) +

  # Add static (non-working/unknown) lights
  geom_point(
    data = subset(blinking_glow_lights, status == "Broken"),
    aes(fill = status),
    shape = 21,
    size = 6,
    color = "gray30",
    stroke = 0.5
  ) +

  # Overlay emoji symbols
  geom_text(
    aes(label = symbol),
    family = "emoji",
    size = 7,
    fontface = "bold",
    color = "red2",
    na.rm = TRUE
  ) +

  # Set colors and labels for legend
  scale_fill_manual(
    values = c(
      "Working" = "yellow",
      "Broken" = "white"
    ),
    name = "Light Status"
  ) +

  coord_fixed() +  # Keep square grid cells

  # Add title and caption
  labs(
    title = "Service Lights",
    caption = "🛠️ Helped     ❓ Unknown"
  ) +

  # Simplify background and layout
  theme_void() +
  theme(
    legend.position = "bottom",
    legend.title = element_text(face = "bold"),
    legend.text = element_text(size = 10),
    legend.key.width = unit(2, "cm"),
    plot.caption = element_text(family = "emoji", size = 10, hjust = 0.5),
    plot.background = element_rect(fill = "white", color = NA),
    plot.margin = margin(20, 20, 60, 20),      # Extra space for legend/caption
    legend.box.margin = margin(t = 10)
  ) +

  # Animate blinking by switching between states
  transition_states(blink_state, wrap = TRUE) +
  ease_aes("linear")

# Save the animation to a GIF (run this manually in Console)
# Do not run during Quarto rendering or it may crash
animate(
  blink_plot,
  fps = 4,                     # Frames per second
  nframes = 8,                 # Total frames
  width = 800,
  height = 600,
  renderer = gifski_renderer("blinking_lights.gif")
)

Total Service Calls

d. Write an artist statement.

This project shows service lights as a grid, using colors to show if they were working, broken, or unknown. It also uses symbols to show if help was needed. I was inspired by lightboards, slot machines, and visualizations that use color, or lack thereof, to represent data in unique ways. I made this using R code, with tools like ggplot, gganimate, and ggfx. The final result is an animated GIF. I cleaned my data and grouped each light by status. I gave each one a color and symbol, then added animation to make the working lights blink and glow.

Problem 3. Statistical Critique

a. Revisit and Summarize

The study includes linear mixed models with the decay constant (k) as the response variable, representing the rate of litter decomposition, and predictor variables including litter type and placement. The authors also use regression analyses with nitrogen and phosphorus concentrations in litter as response variables, and the percentage of litter mass remaining as the predictor variable. These statistical tests compare decomposition rates and nutrient dynamics between buried litter (under gopher mounds) and surface litter. A significant result (p < 0.05) would support the hypothesis that gopher bioturbation accelerates decomposition and nutrient turnover, while also helping to conserve nutrients in fire-prone ecosystems.

Fig. 4

b. Visual clarity

The figure is generally clear. The x-axis represents cumulative litter mass remaining (%) and the y-axis shows nitrogen concentration (%N), with regression lines displayed for each treatment. The authors include R² values for each line, though the actual regression statistics are not provided in the caption—instead, readers are directed to a separate table. Data points and regression lines are overlaid effectively, and the use of faceting by litter type helps isolate trends.

c. Aesthetic clarity

The authors managed visual clutter reasonably well. Faceting helps separate the litter types clearly, and colors are used consistently to distinguish between surface and buried litter. However, some overlapping data points make it difficult to assess point density. A slight increase in transparency could better represent clustering without adding visual noise.

d. Recommendations

The figure is already divided into three panels by leaf type (pine, oak, and mixed), so using both color and shape to represent leaf type within each panel is visually redundant. Meanwhile, surface and buried treatments are only distinguished by slight variations in color shade, which can be difficult to interpret—especially for anyone with color vision impairments.

To improve clarity and accessibility, I recommend keeping the current color scheme but using different shapes to indicate treatment (e.g., circles for surface, triangles for buried). This would make it easier to distinguish between treatments within each panel and to visually compare them across leaf types. As noted in Part 3c, adjusting transparency (alpha) could also help visualize data point density.

Finally, while the figure includes R² values for the regression lines, it might be more helpful to include the other core statistics directly in the caption rather than requiring readers to cross-reference a separate table.