Basic Statistics for CFA Level I — The No-Nonsense, Slightly Snarky Primer

"Data doesn't speak for itself — you have to ask the right questions and not lie with it." — Someone who passed Level I with caffeine

---

Hook: Why stats matter (and why your portfolio cares)

You've already learned Time Value of Money (remember discounting those future cash flows?) and wrestled with the ethics of client recommendations. Basic statistics is the toolkit that connects those ideas: it helps you summarize return histories, estimate future returns, communicate risk honestly, and spot when numbers are being used to mislead. In short: statistics = the language of evidence. Use it well (or ethically suffer the consequences).

---

What this subtopic covers — and why it matters

This mini-lecture walks through:

• Descriptive statistics: summarizing data succinctly
• Probability and distributions: modeling uncertainty
• Sampling & the Central Limit Theorem: turning samples into sensible inferences
• Key business/ethical implications: how misuse misleads investors

All of this feeds into valuation models (remember TVM?), portfolio analytics, and ethical reporting.

---

1) Descriptive statistics — the TL;DR of data

Central tendency

• Mean (arithmetic mean): the average. Great for expected single-period return.
• Geometric mean: the compound annual growth rate — use this when returns are chained over time (TVM vibes!).
• Median: the middle value — robust against outliers.
• Mode: most frequent value — sometimes useful for categorical data.

Why arithmetic vs geometric matters: arithmetic mean overstates multi-period return because it ignores compounding. For a sequence of returns r1, r2, ..., rn:

Arithmetic mean = (r1 + r2 + ... + rn) / n
Geometric mean = ( (1+r1)*(1+r2)*...*(1+rn) )^(1/n) - 1

Use geometric mean when you're thinking like a long-term investor using TVM.

Dispersion (how spread out are returns?)

• Range = max − min
• Interquartile range (IQR) = Q3 − Q1 (robust)
• Variance (σ^2) and Standard deviation (σ) = canonical risk measures

Table quick-compare:

Measure	Sensitive to outliers?	Use when...
Mean	Yes	estimating expected single-period return
Median	No	skewed return distributions
Std Dev	Yes	measuring volatility around mean
IQR	No	summarizing spread without outliers

Shape: skewness & kurtosis

• Skewness tells you asymmetry. Negative skew = left tail (bad big losses).
• Kurtosis tells you tail fatness. High kurtosis = fatter tails = more extreme events than Normal.

Ask: Is the distribution bell-shaped like the Normal? If not, standard risk measures may understate tail risk.

---

2) Probability & distributions — the grammar of uncertainty

• Discrete vs continuous distributions. Binomial vs Normal, basically.
• Normal distribution: handy, mathematically neat, but financial returns often show skew and fat tails.

Important Normal facts:

• About 68% of observations lie within ±1σ of the mean
• About 95% within ±2σ

Z-score transforms: z = (x − μ) / σ — tells you how many standard deviations an observation is from the mean.

Practical finance use: convert asset returns to z-scores to compare different assets on the same risk scale.

---

3) Sampling, Central Limit Theorem (CLT) & standard error

CLT: regardless of the population distribution, the sampling distribution of the sample mean approaches Normal as n increases (usually n≥30 is a handy rule of thumb). This is the statistical miracle that lets you use Normal-based inference even when returns are a bit ugly.

• Standard error (SE) of the mean = σ / sqrt(n) — precision of your sample mean estimate.

Why this matters: when you tell a client "the average return was 8% last decade," you should also know how precisely that 8% is estimated. Small samples = big uncertainty. Ethically relevant? Absolutely — don't pretend precision you don't have.

---

4) Confidence intervals & a pinch of hypothesis testing

Confidence interval for the mean (approx, using Normal or t-distribution):

CI = sample_mean ± (critical_value) * (SE)

Interpretation: a 95% CI means if you repeated the sampling many times, ~95% of CIs would contain the true mean. It does NOT mean there is a 95% chance the true mean falls in this single interval (subtle but important).

Hypothesis testing lets you ask questions like: "Is Fund A truly outperforming benchmark B, or is it just luck?" You'll check whether observed differences are statistically significant given variability and sample size.

---

5) Real-world finance examples & ethical flags

• Expected value of uncertain cash flows: combine probability with TVM. If a payoff has outcomes x_i with probabilities p_i, then expected payoff = Σ p_i * x_i; discount that expected payoff using TVM.

• Using arithmetic mean to present past multi-year returns is misleading; geometric mean (CAGR) aligns with compounding and ethical communication.

• Cherry-picking lookback windows: picking start/end dates that make your strategy look great is statistical malpractice and unethical. You should report sensitivity to time periods and disclose data mining risks.

• Over-reliance on Normal assumptions: downplaying skew and kurtosis can mislead investors about tail risk.

Ask yourself: "Would I be comfortable showing this chart and explanation to a client who asks tough questions?" If not, fix it.

---

Quick cheat-sheet (formulas)

Mean (sample) = x̄ = Σxi / n
Variance (sample) = s^2 = Σ(xi − x̄)^2 / (n − 1)
Std dev (sample) = s = sqrt(s^2)
Geometric mean = (Π(1+ri))^(1/n) − 1
Standard error = s / sqrt(n)
Z-score = (x − μ) / σ

---

Closing: The ethical statistician's checklist

• Use geometric mean for multi-period returns (TVM reminder).
• Always quantify uncertainty (SEs, CIs), not just point estimates.
• Check distributional shape — watch for skewness and fat tails.
• Be transparent about sample size, data selection, and assumptions.

If you're reporting numbers to clients, your job is not to make the numbers look pretty — it's to make them truthful and understandable. That, my friend, is both good practice and ethical practice.

Tags: beginner, humorous, finance