📖 数据可视化方法

Topic: Choosing the Right Chart to Let Data Speak for Itself

1. Introduction: Why Visualization is an Essential Analyst Skill

In L101, we learned how to use frequency distributions and histograms to "bundle" messy data. But histograms are just one tool in the visualization toolkit. In real-world work, different data types and analytical objectives demand different charts.

A scenario: Your portfolio manager asks you to "use one chart to illustrate our portfolio's performance over the past year." What chart would you choose? Pie chart? Bar chart? Line chart?

Choosing the wrong chart = data gets misinterpreted. The CFA Level 1 visualization focus is exactly this: what chart for what data.

2. Core Chart Types and Their Use Cases

2.1 Bar Chart

Definition: Uses rectangular heights to represent values across discrete categories.

Suitable data: Nominal data, ordinal categorical data Typical scenarios: Sector PE valuation comparison, fund size rankings, asset class allocation

⚠️ CFA Trap: Bar charts are for discrete categories; histograms are for continuous data grouped into bins. Bar chart bars have gaps between them; histogram bars have no gaps (continuous).

2.2 Line Chart

Definition: Connects data points across a continuous sequence (usually time) with line segments to show trends.

Typical scenarios: - Stock price charts (the classic line chart) - Fund NAV curve comparison - Macroeconomic indicators (GDP, CPI) time series - Moving average overlays (MA5, MA10, MA20)

⚠️ CFA Trap: Line charts require the X-axis variable to have a natural order (usually time). Using a line chart to connect unordered categories (e.g., different industries) is misleading!

2.3 Scatter Plot

Definition: Uses points in a two-dimensional coordinate system to represent the relationship between two continuous variables.

Reading guide: - Positive correlation: Point cloud runs from bottom-left to top-right - Negative correlation: Point cloud runs from top-left to bottom-right - No correlation: Point cloud appears circular / randomly distributed - Non-linear relationship: Point cloud shows U-shaped, inverted U-shaped, or other curved patterns

Typical scenarios: - Risk vs. return scatter plot (each fund = one point) - Stock Beta vs. expected return - P/E ratio vs. earnings growth rate

Upgrade: Bubble Chart - Adds bubble size as a third variable on top of the scatter plot - Example: X = risk, Y = return, bubble size = fund AUM

2.4 Pie Chart & Donut Chart

Definition: Uses sector angles to represent each part's proportion of the whole.

Usage rules (CFA exam points): 1. Limit categories: More than 5–6 sectors → cluttered and hard to read 2. Sort by size descending: Start from 12 o'clock, proceed clockwise 3. Not for precise comparison: ≈2% vs. ≈3% sectors are nearly indistinguishable by eye 4. Not for trend comparison: Two pie charts side-by-side to compare time periods → inefficient

Typical scenarios: Portfolio asset allocation, sector market cap distribution

⚠️ CFA Trap: Pie charts are unsuitable for showing trends or precise numerical comparisons. When precision is needed, use a bar chart.

2.5 Heat Map

Definition: Encodes numerical values using color intensity; ideal for matrix-format data visualization.

Typical scenarios: - Correlation matrix heat map (the classic application) - Industry × month return matrix - Global market performance heat map

2.6 Box Plot (Box-and-Whisker Plot)

Definition: Summarizes data distribution using five statistics (minimum, Q1, median, Q3, maximum).

   Minimum    Q1   Median   Q3     Maximum
      |———[  |====|  ]———|
         whisker   box   whisker

Core strengths: - A single chart simultaneously shows central tendency, dispersion, skewness, and outliers - Excellent for side-by-side multi-group comparison (e.g., box plots of PE ratios across sectors)

3. Chart Selection Decision Tree

Question 1: What do you want to show?
  ├── Data distribution → Histogram, Box Plot, Frequency Polygon
  ├── Data comparison → Bar Chart (discrete categories)
  ├── Trend over time → Line Chart (time series)
  ├── Variable relationship → Scatter Plot (two continuous variables)
  ├── Composition / proportion → Pie Chart / Donut Chart (whole = 100%)
  └── Matrix pattern → Heat Map

4. Common Visualization Mistakes

5. Real-World Case Study

Scenario: You are an analyst at a fund. Your boss asks you to visualize the following:

1. Annualized return trends of A-shares, H-shares, and US equities over the past 5 years
2. Current portfolio asset allocation breakdown
3. Risk-return scatter distribution of 20 portfolio holdings
4. PE valuation range comparison across industry sectors

What chart should you choose?

6. Key Terminology Quick Reference

7. Practice Questions

Q1 (Concept)

Which chart is most appropriate for showing the monthly NAV change of a fund over the past 36 months? - A. Pie chart - B. Bar chart - C. Line chart - D. Box plot

Q2 (Scenario)

An analyst wants to simultaneously display the median P/E, dispersion, and outliers for 5 different industry sectors. The best choice is: - A. Grouped bar chart - B. Side-by-side box plots - C. Pie chart - D. Scatter plot

Q3 (Trap Identification)

Which of the following statements about data visualization is incorrect? - A. Bar charts are for discrete categories; histograms are for continuous data grouped into bins - B. A pie chart should not have more than 5–6 sectors - C. A bar chart's Y-axis does not need to start at zero, as long as the scale is clearly labeled - D. Scatter plots can reveal non-linear relationships between two variables

Q4 (Terminology)

Compared to a regular scatter plot, what represents the additional dimension in a "bubble chart"? - A. X-axis scale - B. Y-axis scale - C. Point color - D. Point size

8. Answers & Explanations

9. Three Takeaways

Next up L103: Data Cleaning & Wrangling — turning dirty data into analysis-ready gold.