Ch7. Data Visualization and Interpreting Results

Why Visualization Matters: Anscombe’s Quartet

Four datasets with identical descriptive statistics (same mean, variance, correlation) look completely different when plotted. Always visualize your data before modeling.

Chart Selection Guide

Distribution Visualization

Histogram: Frequency distribution of a continuous variable. Reveals shape, center, spread.

Box Plot: Shows Q1, median, Q3, and outliers at a glance.

─○─[  | ]────○
 Min  Q1 Med Q3  Max   ○=outlier

Best for comparing distributions across multiple groups.

Relationship Visualization

Scatter Plot: Relationship between two continuous variables. Reveals correlation, patterns, outliers.

Bubble Chart: Scatter plot with a third variable encoded as bubble size.

Comparison Visualization

Bar Chart: Comparing categorical values. Vertical or horizontal. Stacked Bar: Shows composition and total simultaneously.

Trend Visualization

Line Chart: Changes over time. Essential for time series. Area Chart: Line chart with shaded area beneath.

Correlation Visualization

Heatmap: Correlation matrix with color intensity encoding values. Instant overview of variable relationships.

Pair Plot: Grid of scatter plots for all variable pairs.

Chart selection principles: 1) Start with the question (comparison? distribution? relationship? trend?), 2) Match chart type to data type, 3) Scale chart to data volume. Decoration obscures information — clarity first.

Geographic Visualization

Choropleth Map: Regional values encoded as color intensity (population density, voting rates).

Bubble Map: Bubble sizes encode values at geographic locations.

Dashboard Design Principles

Visual hierarchy: Most important metrics at top-left
Consistency: Same color always means the same thing
White space: Less is more — clarity over density
Interactivity: Filters and drill-downs enable exploration
Context: Show performance vs. target, not just raw numbers

Key Concept Cards

Box Plot ★★★★★ : Q1-median-Q3-outliers in one visual. Essential for comparing group distributions and spotting outliers.

Heatmap ★★★★☆ : Standard for correlation matrix visualization. Color intensity immediately reveals relationships between many variables.

Chart Selection ★★★★★ : Comparison=bar, distribution=histogram/box plot, relationship=scatter, trend=line. Choose by question type.

Practice Quiz

Q1. Comparing income distribution across four cities (Seoul, Busan, Daegu, Gwangju). Best chart?

Box Plot. Shows each city’s median, Q1, Q3, and outliers simultaneously, enabling direct comparison of distributional shape, not just averages. Bar charts only show the mean and miss the full distributional picture.

Q2. Analyzing the relationship between marketing spend and sales revenue. Which chart?

Scatter Plot. Reveals the direction (positive/negative), strength (tight/loose cluster), and shape (linear/curved) of the relationship between two continuous variables. Adding a regression line makes the trend explicit. Potential outliers (unusual spend-to-revenue ratios) are immediately visible.