📖 数据组织综合练习

Topic: From Raw Data to Analytical Insight — End-to-End Data Organization

I. Introduction: What Do You Remember?

From L099 to L103, we covered the full data organization and visualization pipeline:

Objective: Using a realistic financial dataset, connect all the above concepts — from identifying data types through building frequency distributions, selecting visualizations, cleaning outliers, and ultimately deriving meaningful insights.

II. Comprehensive Scenario: Emerging Market ETF Analysis

2.1 Scenario Setup

You are a junior analyst at a fund management company. Your supervisor gives you an emerging market ETF dataset containing the following fields for 50 ETFs:

• Ticker: ETF symbol (e.g., EEM, VWO)
• Category: Morningstar Category (Diversified Emerging Mkts / China Region / India Equity, etc.)
• AUM: Assets Under Management (USD millions)
• ExpenseRatio: Expense ratio (%)
• 3Y_Return: 3-year annualized return (%)
• Morningstar_Rating: Morningstar Rating (1 star to 5 stars)
• Inception_Year: Year of inception

2.2 Sample Data (Simulated)

Ticker    Category              AUM    ExpenseRatio  3Y_Return  Morningstar_Rating  Inception_Year
EEM       Diversified Emerging  23500  0.68          4.2        4                   2003
VWO       Diversified Emerging  82000  0.08          5.1        5                   2005
IEMG      Diversified Emerging  72500  0.09          4.8        4                   2012
FXI       China Region          4500   0.74          -2.5       2                   2004
MCHI      China Region          6200   0.59          -1.8       3                   2011
INDA      India Equity          8900   0.69          7.2        5                   2012
EPI       India Equity          2100   0.85          6.5        3                   2008
EWZ       Brazil Equity         5200   0.63          0.3        3                   2000
...       ...                   ...    ...           ...        ...                 ...
(50 records total)

III. Knowledge Review & Hands-On Practice

3.1 Data Types & Measurement Scales (L099)

Task 1: Identify the data type and measurement scale for each variable

Key Distinction: Why is Morningstar_Rating ordinal rather than interval? - The "quality gap" between 4 stars and 5 stars is not equal to the gap between 1 star and 2 stars - You cannot say "5 stars is 5 times better than 1 star" - Operations on ordinal data are limited to ranking and median; mean and standard deviation are inappropriate

Exam Favorite:

Q: Which measure of central tendency is most appropriate for Morningstar Ratings (1-5 stars)? A: Median, because this is ordinal data — the mean is not meaningful.

3.2 Parameters & Statistics (L100)

Task 2: Calculate the sample mean and sample standard deviation of AUM

Assuming AUM data for 50 ETFs (USD millions, simplified):

23500, 82000, 72500, 4500, 6200, 8900, 2100, 5200, ...

Sample Mean (x̄): $$\bar{x} = \frac{\sum x_i}{n}$$

Assuming ∑AUM = 425,000, n = 50: $$\bar{x}_{AUM} = \frac{425,000}{50} = 8,500 \text{ million}$$

Sample Standard Deviation (s): $$s = \sqrt{\frac{\sum (x_i - \bar{x})^2}{n-1}}$$

Note: The denominator is n-1 (degrees of freedom adjustment) — this makes s an unbiased estimator of σ.

Core Concept Review: - Parameter: A quantity describing a population, e.g., μ (population mean), σ (population standard deviation) - Statistic: A quantity describing a sample, e.g., x̄ (sample mean), s (sample standard deviation) - Sampling Distribution: The probability distribution of a statistic (derived from repeated sampling from the same population)

CFA Level 1 Key Point: Central Limit Theorem — When n is sufficiently large (typically n ≥ 30), the sampling distribution of the sample mean approaches a normal distribution N(μ, σ²/n), regardless of the shape of the original population distribution.

3.3 Frequency Distributions & Contingency Tables (L101)

Task 3: Build a frequency distribution table for Category

Task 4: Contingency Table — Category × Morningstar_Rating

Analytical Findings: - 60% (6/10) of China Region ETFs are rated 1-2 stars — consistent with recent underperformance - 50% (4/8) of India Equity ETFs are rated 4-5 stars — reflecting strong recent returns - Contingency tables reveal relationship patterns between two categorical variables

Exam Focus: Joint Frequency vs. Marginal Frequency - Joint Frequency: Each cell value (e.g., China Region × 1-2 Stars = 6) - Marginal Frequency: Row/column totals (e.g., 1-2 Stars total = 15)

3.4 Data Visualization (L102)

Task 5: Select the correct chart for each analytical purpose

Golden Rules of Visualization: 1. One chart, one core message — Don't overcrowd a single chart 2. Y-axis should start at zero — Unless you deliberately want to exaggerate differences (especially true for bar charts) 3. Clear legends, complete labels — The reader should not need to look back at context 4. Avoid 3D effects — They distort data perception (never use 3D pie charts!)

3.5 Data Cleaning (L103)

Task 6: Identify and handle data problems

Returning to our ETF dataset, suppose you discover the following issues:

Problem A: Missing Values

• 3 ETFs have no Morningstar_Rating (new funds, not yet rated)
• Solution: Use pairwise deletion for analyses involving ratings; for coverage reporting, label as "NR" without deletion

Problem B: Outliers

• One ETF shows AUM of 1,250,000 (125 trillion!)
• After verification: should be 12,500 (an extra zero)
• Solution: Data correction (correct to 12,500)

Problem C: Duplicates

• VWO appears twice
• Both records are identical → Deduplicate (delete one)

Problem D: Inconsistent Formats

• Some Inception_Year entries are 4-digit years (2003), others are full dates (2003-04-01)
• Solution: Standardize — Extract year uniformly, convert to numeric

Common Data Problems Quick Reference

IV. Comprehensive Practice Questions

Multiple Choice

Q1: Which measure of central tendency is most appropriate for Morningstar Ratings (1-5 stars)? - A. Arithmetic mean - B. Median - C. Geometric mean - D. Weighted mean

Q2: To analyze the relationship between AUM and ExpenseRatio, the most appropriate chart is a: - A. Bar chart - B. Histogram - C. Scatter plot - D. Pie chart

Q3: The denominator in the sample standard deviation formula is n-1. This reflects: - A. Loss of one degree of freedom, making s an unbiased estimator of σ - B. Precision limitations in data collection - C. A special rule by CFA Institute - D. The larger the sample, the more n-1 is needed

Q4: Which of the following operations is "data cleaning" rather than "data wrangling"? - A. Merging two datasets by ticker - B. Flagging a CAGR of 25,000% as an outlier - C. Grouping by Category to calculate average return - D. Sorting data by AUM in descending order

Q5: In a contingency table, the cell showing 15 (Diversified × 4-5 Stars) represents a: - A. Marginal frequency - B. Joint frequency - C. Conditional frequency - D. Cumulative frequency

Answers

Q1: B — For ordinal data, the median is the appropriate measure of central tendency. The mean is not meaningful.

Q2: C — A scatter plot is ideal for examining the relationship between two continuous variables (possibly adding a regression line).

Q3: A — The denominator n-1 accounts for using x̄ (a sample statistic) in place of μ (population parameter), losing one degree of freedom. This makes s² an unbiased estimator of σ².

Q4: B — Outlier detection and handling is data cleaning (correcting data errors). A (merging), C (aggregation), and D (sorting) are data wrangling operations.

Q5: B — A cell at the intersection of a row and column is a joint frequency. Marginal frequencies are row/column totals; conditional frequencies are proportions within a row or column.

V. Module Knowledge Map

Module 2.2: Organizing & Visualizing Data (L099-L104)

L099 Data Types ──── Quantitative vs. Qualitative, Four Scales
    │
L100 Parameters/Stats ── μ vs. x̄, CLT, Sampling Distributions
    │
L101 Frequency Distributions ── Tables, Contingency Tables, Joint/Marginal Freq
    │
L102 Data Visualization ── Bar/Histogram/Scatter/Box/Pie Charts
    │
L103 Data Cleaning ── Missing Values/Outliers/Duplicates/Format Issues
    │
L104 Comprehensive Practice ◀── You are here!

VI. Exam Priorities

Next Module Preview: Starting from L105, we enter Module 2.3: Descriptive Statistics — measures of central tendency, dispersion, skewness, and kurtosis. This is one of the most important and highest-density modules in CFA Level 1 Quantitative Methods. Get your calculator ready!

CFA Level 1 · Quantitative Methods · Module 2.2 Complete | L104 Comprehensive Practice · 2026-07-10