Chapter 14 Exercises: The Grammar of Graphics

How to use these exercises: This chapter is conceptual — no code required. You'll think, sketch, analyze, and critique. Many exercises ask you to draw or describe charts on paper. This is intentional: designing charts is a thinking activity, and paper is the best medium for thinking. Work through the sections in order; each section builds on the previous one.

Difficulty key: (star) Foundational | (star)(star) Intermediate | (star)(star)(star) Advanced | (star)(star)(star)(star) Extension

Part A: Grammar of Graphics Components (star)

These exercises test whether you can identify and explain the six components of the grammar of graphics.

Exercise 14.1 — Identifying components

For each of the following chart descriptions, identify the six grammar of graphics components (data, aesthetic mappings, geometric object, scale, coordinate system, faceting):

(a) A bar chart showing average life expectancy by continent, with bars colored by continent.

(b) A scatter plot showing the relationship between a country's GDP per capita (x-axis) and its CO2 emissions per capita (y-axis), with point size representing population.

(c) A set of six small histograms, one per WHO region, each showing the distribution of vaccination rates within that region.

Guidance **(a)** - Data: Country-level dataset, aggregated to continent means - Aesthetics: x = continent, y = mean life expectancy, color = continent - Geom: Bar - Scale: y-axis linear starting at 0; categorical x-axis; distinct colors per continent - Coordinates: Cartesian - Faceting: None **(b)** - Data: Country-level dataset, one row per country - Aesthetics: x = GDP per capita, y = CO2 per capita, size = population - Geom: Point (circle) - Scale: x and y linear (or log for GDP); size proportional to population - Coordinates: Cartesian - Faceting: None **(c)** - Data: Country-level vaccination rates - Aesthetics: x = vaccination rate (binned), y = count - Geom: Bar (histogram bars) - Scale: x-axis linear; y-axis linear (count) - Coordinates: Cartesian - Faceting: By WHO region (6 panels)

Exercise 14.2 — Changing one component

Start with this chart description: "A scatter plot of GDP per capita (x) vs. vaccination rate (y) for all countries, colored by income group."

Describe what the chart would look like if you changed only one component at a time:

(a) Change the geom from points to lines (connecting countries in alphabetical order).

(b) Change the coordinate system from Cartesian to polar.

(c) Add faceting by WHO region.

(d) Change the color aesthetic from income group to vaccination rate (a continuous variable).

For each, explain whether the change produces a useful chart or a confusing one, and why.

Guidance **(a)** Lines connecting countries alphabetically would be confusing — there's no meaningful ordering from "Afghanistan" to "Zimbabwe." The connecting lines imply continuity that doesn't exist. This is a *bad* change. **(b)** Polar coordinates would wrap the x-axis into a circle, making the scatter plot nearly unreadable. GDP values would map to angles, which is not intuitive. This is a *bad* change. **(c)** Faceting by region would create one scatter plot per region, making it easier to see the GDP-vaccination relationship within each region without the visual clutter of all regions on one chart. This is a *good* change. **(d)** Changing color from a categorical variable (income group) to a continuous variable (vaccination rate) would create a gradient effect. But vaccination rate is already on the y-axis, so the color would be redundant — points higher on the chart would be darker. This is *redundant but not harmful* — sometimes called "double encoding." It can reinforce the message but doesn't add new information.

Exercise 14.3 — Grammar decomposition

Find a chart in a news article, textbook, or website (not from this book). Decompose it into its grammar of graphics components. Write out all six components. Then answer: Is there any component the chart handles particularly well? Any component that could be improved?

Guidance This is an open-ended exercise. A strong answer clearly identifies each component and offers a specific critique — not just "it looks good" or "it's confusing" but identifying exactly which component contributes to the strength or weakness. For example: "The scale on the y-axis starts at 50 rather than 0, which visually exaggerates the differences between bars" (a scale critique) or "The use of faceting by year makes the trend over time clear without cluttering a single panel" (a faceting compliment).

Part B: Chart Type Selection (star)(star)

These exercises test your ability to match analytical questions to appropriate chart types.

Exercise 14.4 — Choose the chart

For each question below, recommend the most appropriate chart type and justify your choice. Also name one chart type that would be a poor choice and explain why.

(a) "How do math test scores compare across five school districts?"

(b) "Is there a relationship between hours of sleep and exam performance among 200 college students?"

(c) "What does the distribution of home prices look like in our city?"

(d) "How has the unemployment rate changed from 2000 to 2023?"

(e) "What percentage of our company's revenue comes from each product line?"

Guidance **(a)** Best: Bar chart (or box plot if showing distribution within each district). Poor: Line chart — there's no meaningful ordering from district A to district E, so connecting lines imply false continuity. **(b)** Best: Scatter plot — two continuous variables and a relationship question. Poor: Pie chart — there are no parts of a whole here, and two continuous variables can't be shown in a pie. **(c)** Best: Histogram — one continuous variable, distribution question. Poor: Bar chart — home prices are continuous, not categorical. (A bar chart of price ranges is essentially a histogram, but creating arbitrary categories misses the point.) **(d)** Best: Line chart — one variable changing over time. Poor: Scatter plot without connecting lines — it would show the same data but miss the temporal continuity that makes trends visible. **(e)** Best: Horizontal bar chart (or stacked bar chart, or even pie chart if there are only 3-4 categories). Poor: Line chart — there's no sequential ordering across product lines.

Exercise 14.5 — The "it depends" exercise

For each scenario, two chart types could work. Explain under what circumstances you would choose each one.

(a) Comparing vaccination rates across 6 regions: bar chart vs. dot plot.

(b) Showing test score distributions across 4 classes: histogram vs. box plot.

(c) Showing how revenue breaks down across product lines: bar chart vs. pie chart.

Guidance **(a)** Bar chart: when the audience is general and expects a familiar format; when the values are easy to compare and you want the visual weight of bars. Dot plot: when you want to emphasize precise values over visual weight; when you have many categories and bars would take up too much space; when you want to minimize non-data ink (dots have less "ink" than bars). **(b)** Histogram: when you want to see the full shape of each distribution (bimodality, skewness, gaps). But overlapping histograms for 4 classes gets cluttered. Box plot: when you want a clean, compact comparison of medians, spreads, and outliers across groups, sacrificing distributional detail for clarity. **(c)** Bar chart: almost always the better choice because length is easier to judge than angle. Pie chart: when there are only 2-3 categories, one is clearly dominant, and you want to emphasize the "part of a whole" framing (e.g., "Product A accounts for two-thirds of our revenue").

Exercise 14.6 — Marcus's menu analysis

Marcus wants to analyze his coffee shop sales data. He has the following questions. For each, recommend a chart type and sketch a rough version (labeled axes, approximate marks — it does not need to be beautiful).

(a) "Which product category (Coffee, Tea, Food, Other) has the highest total revenue?"

(b) "How does daily revenue change over the course of a year?"

(c) "Is there a relationship between the time of day and the average transaction amount?"

(d) "What does the distribution of individual transaction amounts look like?"

Guidance **(a)** Bar chart: x = category, y = total revenue. Four bars, sorted by height (tallest first) for easy comparison. **(b)** Line chart: x = date, y = daily revenue. A single line showing the trend, possibly with a smoothed overlay to see the seasonal pattern. **(c)** Bar chart: x = time of day (Morning, Afternoon, Evening), y = mean transaction amount. Only 3 categories. Alternatively, a box plot if you want to show the spread within each time period. **(d)** Histogram: x = transaction amount (binned), y = count. This will show whether most transactions are small with a few large ones (right-skewed, likely) or roughly symmetric.

Part C: Sketching and Planning (star)(star)

These exercises require you to sketch charts on paper (or describe them in enough detail that someone else could draw them).

Exercise 14.7 — The chart plan

Using the chart plan template from Section 14.8, create a complete chart plan for the following analytical question:

"Do countries with higher healthcare spending per capita tend to have higher vaccination rates?"

Your plan should specify: question, chart type, data source, x-axis, y-axis, color encoding (if any), faceting (if any), title, annotations, and audience.

Guidance A strong chart plan might look like: - Question: Is there a positive relationship between healthcare spending and vaccination rates? - Chart type: Scatter plot - Data: Country-level dataset, one row per country, most recent year - x-axis: Healthcare spending per capita (USD), labeled "Healthcare Spending per Capita ($)" - y-axis: Vaccination rate (%), labeled "Vaccination Rate (%)" - Color: WHO region (categorical palette, 6 colors) - Facets: None (or optionally faceted by income group) - Title: "Higher Healthcare Spending Is Associated With — But Does Not Guarantee — Higher Vaccination Rates" - Annotations: Label notable outliers (e.g., countries with high spending but low rates); add a trend line - Audience: Explanatory — for a policy report

Exercise 14.8 — Three views of one dataset

Imagine you have a dataset of 500 employees at a company with columns: department, salary, years_of_experience, gender, and performance_rating.

Sketch three different charts using this dataset, each answering a different question. For each chart, state: (a) the question, (b) the chart type, (c) the aesthetic mappings.

Guidance Example answers (many valid options exist): Chart 1: "How do average salaries compare across departments?" Bar chart. x = department, y = mean salary, color = single color. Chart 2: "Is there a relationship between years of experience and salary?" Scatter plot. x = years_of_experience, y = salary, color = department (or gender). Chart 3: "What is the distribution of performance ratings?" Histogram. x = performance_rating (binned), y = count.

Exercise 14.9 — Redesign a default chart

Imagine a plotting library produces the following default bar chart of average test scores across four schools: colored bars with a dark gray background, heavy black gridlines, a border around the plot, a 3D bevel effect on each bar, and a title that says "Bar Chart."

List at least six specific changes you would make to improve this chart, citing Tufte's principles where relevant.

Guidance 1. Remove the 3D bevel effect — it's chartjunk that distorts bar heights (Tufte: eliminate chartjunk) 2. Change the background from dark gray to white — the background is non-data ink (Tufte: maximize data-ink ratio) 3. Change the gridlines from heavy black to faint gray — gridlines should support, not compete with, the data 4. Remove the border — the border is non-data ink 5. Change the title from "Bar Chart" to something descriptive: "Lincoln Elementary Leads District in Math Scores" — the title should state a finding, not a chart type 6. Add axis labels with units (e.g., "Average Score (out of 100)") 7. Ensure the y-axis starts at zero 8. Consider removing the legend if color doesn't encode a variable (if all bars are different colors just for aesthetics, make them all the same color)

Exercise 14.10 — Progressive project chart plans

For the progressive project (vaccination rate analysis), create chart plans for at least three visualizations you plan to build in Chapter 15. Each plan should follow the template from Section 14.8. Your three charts should: - Use at least two different chart types - Include at least one exploratory and one explanatory chart - Address different research questions about the vaccination data

Guidance This is open-ended but should demonstrate thoughtful chart selection. A strong answer might include: 1. An exploratory histogram of vaccination rates to understand the distribution shape 2. An explanatory bar chart comparing WHO regions to highlight disparities 3. An explanatory line chart showing vaccination rate trends over time for selected countries Each plan should have complete specifications including question, chart type, axis mappings, and audience.

Part D: Critique and Analysis (star)(star)(star)

These exercises develop your ability to evaluate charts critically.

Exercise 14.11 — Spot the manipulation

For each of the following chart descriptions, identify the specific misleading technique and explain how to fix it.

(a) A bar chart comparing Company A's revenue ($4.2 billion) to Company B's revenue ($3.8 billion). The y-axis runs from $3.5 billion to $4.5 billion.

(b) A line chart titled "Crime Is Skyrocketing!" that shows crime rates from 2019 to 2023. Crime rates actually peaked in 2020 and have been declining since, but the chart starts at the trough in 2019.

(c) An infographic showing that Country X spends "twice as much" on education as Country Y, illustrated by a schoolhouse icon for X that is twice as tall and twice as wide as the icon for Y.

(d) A chart with two y-axes showing ice cream sales (left axis) and drowning incidents (right axis) over the same time period, with the axes scaled so the two lines track each other closely.

Guidance **(a)** Truncated y-axis. The axis starts at $3.5B, making the $0.4B difference (about 10%) look like Company A's bar is nearly twice the height of Company B's. Fix: start the y-axis at zero. **(b)** Cherry-picked time range. Starting at a trough makes the subsequent peak look like a sustained increase. A longer time range (e.g., 2005-2023) would show the fuller picture, likely revealing that the 2020 spike was an anomaly in a longer-term trend. **(c)** Area distortion. Doubling both height and width makes the larger schoolhouse icon four times the area (2x * 2x = 4x), visually suggesting a four-fold difference when the actual difference is two-fold. Fix: use a bar chart, or scale the icons by only one dimension. **(d)** Dual y-axes with manipulated scales. By independently choosing the scale for each axis, the creator can make any two lines appear to track each other. This suggests a causal relationship between ice cream and drowning (in reality, both are driven by a confounding variable: warm weather). Fix: remove dual axes; if comparison is needed, use separate panels with clearly labeled axes.

Exercise 14.12 — The Tufte audit

Find a chart online or in a publication. Perform a "Tufte audit":

  1. Estimate the data-ink ratio (roughly — high, medium, or low).
  2. Identify any chartjunk and list it.
  3. Identify any non-data ink that could be removed.
  4. Suggest specific changes that would increase the data-ink ratio.
Guidance A strong answer is specific. Instead of "the chart has too much clutter," identify exactly what: "The chart has a dark gray background, heavy gridlines, a border, 3D effects on the bars, and a legend box with its own border. Removing the background, reducing gridline weight, removing the border, eliminating 3D effects, and removing the legend border would increase the data-ink ratio from approximately low-medium to high."

Exercise 14.13 — Exploratory vs. explanatory

For each of the following situations, determine whether the visualization should be exploratory or explanatory, and list three specific design choices you would make differently depending on your answer.

(a) You've just loaded a new dataset and want to understand what's in it.

(b) You're preparing a slide for a board of directors meeting showing quarterly sales trends.

(c) You're checking whether the residuals of your regression model are normally distributed.

(d) You're creating a figure for a peer-reviewed journal article.

Guidance **(a)** Exploratory. Design choices: default colors (no time polishing palette), no title (you know what you're looking at), no annotations, run quickly through many variables. **(b)** Explanatory. Design choices: descriptive title stating the finding, brand-consistent colors, clean and minimal (the board has 30 seconds to read it), key values annotated directly on the chart. **(c)** Exploratory. Design choices: default histogram or Q-Q plot, no title needed, no audience considerations — you're just checking an assumption. **(d)** Explanatory. Design choices: publication-quality resolution (300+ DPI), clear axis labels with units, grayscale or colorblind-safe palette (for print compatibility), precise legend, descriptive caption.

Part E: Deeper Thinking (star)(star)(star)

These exercises push toward synthesis and critical evaluation.

Exercise 14.14 — When the grammar breaks down

The grammar of graphics is powerful but not universal. Describe a type of visualization that doesn't fit neatly into the grammar's framework (data + aesthetics + geoms + scales + coordinates + facets). Why does it resist decomposition? How might you extend the grammar to accommodate it?

Guidance Examples of visualizations that challenge the standard grammar: network diagrams (nodes and edges don't map cleanly to x/y aesthetics), word clouds (position is often arbitrary, size encodes frequency but spatial arrangement isn't meaningful), and geographic maps with overlaid flow arrows (combining geographic coordinates with origin-destination pairs). The grammar could be extended by adding new geom types (network edges, text-with-size), new coordinate systems, or new aesthetic channels. Wilkinson's original book actually addresses some of these extensions.

Exercise 14.15 — The ethics of emphasis

You are a data analyst preparing a report on your company's diversity hiring. The data shows that the percentage of underrepresented minorities hired increased from 12% to 14% over three years.

(a) Design a chart that honestly but favorably presents this data (emphasizing progress).

(b) Design a chart that honestly but unfavorably presents this data (emphasizing how far there is to go).

(c) Discuss: Both charts use the same accurate data. Is either one "wrong"? What ethical principles should guide your choice?

Guidance **(a)** A line chart of the percentage over time, with the y-axis from 10% to 16%, titled "Diversity Hiring Improves for Third Consecutive Year." The narrower axis range makes the increase visually dramatic. **(b)** A bar chart showing 14% hired vs. an industry target of 25%, with the y-axis from 0% to 30%, titled "Diversity Hiring Remains Well Below Industry Targets." The wide axis and benchmark make the gap obvious. **(c)** Neither chart lies — both use accurate data. But they frame the same truth differently. The ethical question is about context and intent: Are you providing full context or selectively framing? Is the audience getting the information they need to make an informed judgment? A complete report might include both views, or a single chart that shows both the trend (progress) and the benchmark (gap remaining).

Exercise 14.16 — Design for a different audience

You have data showing that air pollution levels in a city decreased by 15% after a new transit system was built. Design (describe or sketch) three different charts presenting this finding, each for a different audience:

(a) City council members deciding on future transit investments

(b) A community newsletter read by residents of affected neighborhoods

(c) An academic journal on urban planning

Explain how your design choices differ for each audience and why.

Guidance Key differences might include: level of detail (city council wants bottom-line impact and cost context; academics want methodology and confidence intervals; residents want personal relevance), visual complexity (council may need a simple bar chart they can grasp in 10 seconds; academics expect scatter plots or time series with error bars), and framing (residents care about health implications; council cares about ROI; academics care about causality vs. correlation).

Part F: Synthesis and Application (star)(star)(star)(star)

Exercise 14.17 — The visualization taxonomy

Create a visual taxonomy (a flowchart or decision tree, drawn on paper or described in text) that helps someone choose a chart type based on three questions: (1) What type of question am I asking? (2) How many variables am I plotting? (3) What data types are those variables (categorical or continuous)?

Your taxonomy should cover at least 6 chart types and produce a specific recommendation for each leaf node.

Guidance A strong taxonomy starts with the question type (comparison, relationship, distribution, composition, trend) then branches on the number and types of variables. For example: Relationship -> Two variables -> Both continuous -> Scatter plot. Comparison -> One categorical + one continuous -> Bar chart. Distribution -> One continuous variable -> Histogram. The key is that the logic is clear and the recommendations are justified.

Exercise 14.18 — Anscombe's lesson revisited

Anscombe's Quartet (1973) demonstrated that four datasets with identical summary statistics can look very different when plotted. More recently (2017), researchers Justin Matejka and George Fitzmaurice at Autodesk created the "Datasaurus Dozen" — thirteen datasets with identical summary statistics that form shapes including a dinosaur, a star, and an X pattern when plotted.

(a) Without looking up the Datasaurus Dozen, explain why summary statistics (mean, standard deviation, correlation) fail to capture the structure of these datasets.

(b) What specific visual feature does each dataset in Anscombe's Quartet illustrate that the summary statistics miss?

(c) How does this argument support the claim that "visualization is a form of thinking, not just a communication tool"?

Guidance **(a)** Summary statistics compress the entire distribution into a few numbers, losing information about shape, clusters, outliers, and nonlinear patterns. Mean and standard deviation describe a normal distribution well, but they say nothing about whether the data is actually normal. **(b)** Dataset I: a linear relationship. Dataset II: a nonlinear (curved) relationship. Dataset III: a perfect linear relationship with a single outlier shifting the regression. Dataset IV: no relationship except one influential outlier creating a false correlation. The correlation coefficient is the same for all four, but the underlying patterns are completely different. **(c)** Visualization reveals structure that numbers hide. When you plot your data, you're not just making it "pretty" for an audience — you're discovering properties of the data that no summary statistic captures. This makes visualization an analytical tool, not just a presentation tool.

Exercise 14.19 — Cross-cultural chart design

Research suggests that reading direction (left-to-right vs. right-to-left) can affect how people interpret time-based charts. In cultures that read left-to-right, time is naturally perceived as flowing from left to right on a horizontal axis. In cultures that read right-to-left, the opposite may be true.

Discuss: How might this affect the design of a line chart showing vaccination trends over time if your audience includes people from both reading directions? What design choices could help make the chart accessible to everyone?

Guidance This is an open-ended research question. Strong answers might suggest: clear axis labels with years explicitly marked (so the direction is unambiguous regardless of reading convention), arrows or annotations indicating the direction of time, thoughtful placement of "start" and "end" labels, or using a vertical time axis (top-to-bottom) which may be more culturally neutral. The broader point is that design is not culturally neutral — assumptions embedded in chart conventions may not transfer across audiences.

Part G: Progressive Project (star)(star)

Exercise 14.20 — Your vaccination data chart plans

Create a "Chart Plan Document" for the progressive project. This document should contain at least five chart plans (using the template from Section 14.8), each addressing a different research question about the WHO vaccination dataset. Your plans should:

  1. Include at least three different chart types
  2. Include at least one chart with color encoding
  3. Include at least one chart with faceting
  4. Include at least two explanatory charts with descriptive titles
  5. Include at least one exploratory chart

This document will guide your work in Chapters 15-17.

Guidance A complete answer has five detailed chart plans with all template fields filled in. Example plans: 1. Histogram of vaccination rates across all countries (exploratory, distribution question) 2. Bar chart of mean vaccination rates by WHO region in 2023 (explanatory, comparison) 3. Line chart of vaccination rates over time for five selected countries (explanatory, trend) 4. Scatter plot of GDP per capita vs. vaccination rate, colored by income group (relationship) 5. Faceted histograms of vaccination rate distribution, one panel per income group (distribution comparison)

Exercise 14.21 — Critique a real public health chart

Find a published chart related to public health (vaccination rates, disease incidence, health spending — any topic). This could be from the WHO, CDC, Our World in Data, a news article, or a journal paper.

(a) Identify the grammar of graphics components used. (b) Evaluate the chart using Tufte's principles (data-ink ratio, chartjunk). (c) Is it exploratory or explanatory? (d) Suggest at least two specific improvements.

Guidance This exercise bridges the chapter's concepts with real-world practice. Strong answers include the specific source, a detailed component breakdown, and actionable improvement suggestions. For example: "The WHO's dashboard uses a choropleth map (geom: filled polygon, aesthetic: color = vaccination rate, coordinates: geographic). The data-ink ratio is high — minimal borders and gridlines. However, the sequential color palette uses red-to-green, which is not colorblind-safe. Improvement: use a blue-to-yellow palette. Additionally, the legend shows continuous values but uses only 5 discrete color bins, which loses precision. Improvement: use a continuous color gradient."

Part H: Discussion and Debate (star)(star)(star)

Exercise 14.22 — The pie chart debate

Write a one-paragraph argument FOR the use of pie charts, and a one-paragraph argument AGAINST. Then state your own position and defend it.

Guidance FOR: Pie charts are universally recognized, immediately communicate "part of a whole," and work well for 2-3 categories where one dominates. They are familiar to non-technical audiences and sometimes the communication goal is a rough sense of proportion, not precise comparison. AGAINST: Cleveland and McGill's research shows that humans judge angles less accurately than lengths. Pie charts with more than 3-4 slices become unreadable. They cannot show negative values or values that don't sum to a meaningful whole. A bar chart conveys the same information more accurately in every case. A strong personal position acknowledges both sides and articulates a specific standard (e.g., "I use pie charts only when I have 2-3 categories and the audience expects them").

Exercise 14.23 — Is "data-ink ratio" always right?

Tufte's data-ink ratio principle says to maximize the proportion of ink devoted to data. But some research (e.g., work by Scott Bateman and colleagues at the University of Saskatchewan) suggests that charts with decorative elements ("chart junk") can be more memorable and engaging than minimalist charts.

(a) When might Tufte's minimalism be the right approach?

(b) When might some decoration or ornamentation be acceptable or even beneficial?

(c) Does the answer depend on the audience? The medium (print vs. screen)? The purpose (analysis vs. communication)?

Guidance **(a)** For analytical, scientific, and technical audiences where precision matters: academic papers, financial reports, medical dashboards. When the audience needs to make decisions based on exact values, minimalism is essential. **(b)** For general public communication where memorability matters more than precision: infographics, public health campaigns, newspaper articles. Bateman et al.'s 2010 study found that "embellished" charts were remembered more accurately after a two-to-three-week delay. **(c)** Yes — audience, medium, and purpose all matter. An infographic on social media might benefit from decoration that grabs attention. A figure in a journal article should not. The key is intentionality: every design element should serve a purpose, whether that purpose is precision or engagement.

Exercise 14.24 — Visualization and trust

A pharmaceutical company publishes a chart showing that their drug reduces symptom severity by 50%. The chart shows the treatment group's average score dropping from 8.0 to 4.0 on a 10-point scale, while the placebo group drops from 8.0 to 7.5.

(a) What information is missing from this description that you would need to evaluate the chart?

(b) How could the chart be designed to be maximally informative and honest?

(c) How could the same accurate data be presented in a way that overstates the drug's effectiveness?

Guidance **(a)** Missing information: sample sizes, confidence intervals or error bars, statistical significance, the definition of "symptom severity," the time period, whether the study was randomized and blinded, potential side effects. **(b)** Show both groups with error bars/confidence intervals. Include sample sizes in the labels. Use a title that states the result neutrally: "Treatment Group Shows Greater Symptom Reduction Than Placebo (N=200, p<0.01)." Show individual data points or distributions, not just means. **(c)** Show only the treatment group (omitting the placebo context). Use a truncated y-axis (e.g., 3 to 9) to make the drop look more dramatic. Use a 3D effect. Title it "Revolutionary Drug Cuts Symptoms in Half!"

Exercise 14.25 — Your visualization philosophy

Based on everything you learned in this chapter, write a short personal "visualization philosophy" — 3-5 principles that will guide your chart-making throughout this course and beyond. These should be specific enough to be actionable (not just "make good charts") and should reflect the concepts from this chapter.

Guidance Example principles (students should develop their own): 1. Always start with a question, not a chart type. The question determines the chart, not the other way around. 2. Sketch on paper first. If I can't draw it by hand, I don't understand what I'm trying to show. 3. Start exploratory, then polish to explanatory. Don't spend 30 minutes on axis labels for a chart I'm going to throw away. 4. Every chart gets a descriptive title that states the finding, not just the topic. 5. If in doubt, use a bar chart. It's almost never the wrong choice for a comparison, and it's almost always clearer than a pie chart.

End of Chapter 14 Exercises