Untitled Flashcards Set

📚 WEEK 5 STUDY GUIDE

Topic: Data Visualization – How to Read, Interpret, and Critique Graphs

🧠 Lecture Content

1. Identifying Features of Graphs

To correctly interpret a graph, always check:

Title: Should clearly state what the graph is about.

Axes:

X-axis (horizontal) and Y-axis (vertical) must have:

Labels indicating the variable

Units (if applicable)

Appropriate and consistent scales

Legend: Crucial when comparing groups or multiple variables. It must be:

Clear, consistent, and not misleading.

Data Display: The core of the graph—could be:

Bars, points, lines, dots, or shaded areas (like error bands).

Error Bars/Bands:

Indicate variability or uncertainty around a mean.

Wider bars = greater uncertainty; Narrower bars = more confidence.

Color and Clarity:

Choose colors logically and accessibly.

Avoid clutter (too many labels, lines, or unnecessary design elements).

2. Different Types of Graphs (When and Why to Use Them)

Pie Charts

Use when showing parts of a whole (e.g., percentage of water usage by appliance).

Must add to 100%.

Avoid when:

Too many categories

Categories are not mutually exclusive

Labels are missing or hard to interpret

Bar Plots

Compare values across categorical variables.

Can be vertical or horizontal.

When to use:

Comparing survey responses by category (e.g., COVID-19 cases by race).

Best practices:

Use error bars for averages to show variability.

Avoid redundant or cluttered labels.

Sometimes better to show individual data points (especially with small samples).

Histograms

Specialized bar plots for frequency distributions.

X-axis = continuous variable, Y-axis = frequency.

Bin width matters:

Too wide = oversimplifies data

Too narrow = creates noise

Histograms vs. Bar Plots:

Histogram: continuous x-axis, frequency on y-axis

Bar plot: categorical x-axis, value on y-axis

Line Graphs

Best for continuous data (e.g., time series).

Connect data only when sequential.

Don’t connect missing data—doing so misrepresents trends.

Can have two y-axes—but only with clear labeling.

Use error bands to indicate variability over time.

Scatter Plots

Show relationships between two continuous variables.

Display every individual data point.

Great for identifying trends, clusters, or outliers.

Can become overwhelming with too much data.

Consider transparency or jittering to reduce overlap.

3. How to Critique Graphs (What Makes Graphs Misleading or Confusing)

❌ Inverted Axes

Common trick to reverse the message of the data.

Readers expect upward = increase, downward = decrease.

❌ Changing Axis Scales or Legends Midway

Makes comparisons impossible or misleading.

Example: Y-axis starts at 0 in one part, starts at 50 in another.

❌ Truncated Axes

Cutting off part of the Y-axis can exaggerate or minimize differences.

OK only when justified (e.g., focusing on relative change).

Best when:

Range is narrow and meaningful

You clearly note the truncation

❌ Inconsistent or Misleading Color Use

Color should represent data meaningfully:

Sequential data: one color, light to dark (e.g., light blue to dark blue)

Divergent data: two opposing colors (e.g., red for negative, blue for positive)

Categorical data: different distinct colors

Avoid red-green contrasts due to colorblind accessibility issues.

❌ Distorted Aspect Ratio

Stretching or compressing axes can exaggerate or hide trends.

Keep aspect ratios that maintain visual accuracy of slopes or change.

❌ Clutter or Redundancy

Over-labeling, unnecessary 3D effects, or extra text can confuse viewers.

Simplify without losing important data:

Don’t sacrifice clarity for aesthetics.

✅ Include All Necessary Information

Axes with units and labels

Title and legend (if needed)

Annotations or explanations if something unusual is shown

Use simple 2D graphics unless 3D adds real value

📘 Readings

📖 Pandey et al. (Deceptive Visualizations)

Focus: How graphs deceive

Key Deceptive Tactics:

Truncated axes exaggerating effects

Selective data reporting (e.g., omitting inconvenient data points)

Misleading color gradients

Overuse of complexity to overwhelm or mislead

Purpose of reading:

Recognize and analyze deceptive practices

You DO NOT need to remember experimental details or study methods

📖 Kelleher et al. (Ten Guidelines for Good Visualization)

Focus: Principles for designing effective visualizations

General Themes:

Use appropriate graph types for your data

Avoid “chartjunk” (unnecessary visual clutter)

Ensure readability (label size, spacing, clarity)

Design for accessibility (colorblind-friendly palettes, font sizes)

Display uncertainty when relevant (error bars, confidence intervals)

Keep it simple and interpretable

Purpose of reading:

Learn to spot good vs bad graphs

You DON’T need to memorize the full list of 10