STATISTICS 110: PROBABILITY

Lectures 7 & 8 — Random Variables, Distributions, Gambler's Ruin

Harvard University | Prof. Joe Blitzstein | Complete University-Level Notes

📺 Lecture 7 📺 Lecture 8

• Gambler's Ruin Problem • Random Variables (formal)

• First Step Analysis • PMF (Probability Mass Function)
• Random Variables (intro) • CDF (Cumulative Distribution Function)
• Bernoulli Distribution • Hypergeometric Distribution
• Binomial Distribution • Independence of R.V.s

LECTURE 7 — PART 1: GAMBLER'S RUIN PROBLEM

The Gambler's Ruin is a classic probability problem that introduces first step analysis — a powerful technique for
solving recursive probability problems.

Problem Setup
📘 Definition: Gambler's Ruin Problem
A gambler starts with $i (where 0 < i < N).
At each step: win $1 with probability p, or lose $1 with probability q = 1 − p.
The game ends when the gambler reaches $0 (ruin) or $N (target/goal).
Question: What is the probability of reaching $N before hitting $0?

Setting Up the Equation — First Step Analysis

💡 Note: What is First Step Analysis?
First Step Analysis means: condition on what happens at the VERY FIRST step.
This converts a complex problem into a manageable recursive equation.
Strategy: P(win from state i) = P(win|win first step)·P(win first step) + P(win|lose first step)·P(lose first
step)

Let pᵢ = P(reach $N | start with $i)

Boundary conditions (known values):

• p₀ = 0 (already ruined — impossible to reach $N)
• pₙ = 1 (already at goal — already won)

For 0 < i < N, applying First Step Analysis:

🔢 Formula: Gambler's Ruin Recurrence
pᵢ = p · pᵢ₊₁ + q · pᵢ₋₁
 Meaning: Starting from $i, with prob p we move to $i+1, with prob q we move to $i−1.
This is a linear second-order recurrence relation.

Solving the Recurrence

📐 Derivation: Gambler's Ruin Solution
Rewrite recurrence: p·pᵢ₊₁ − pᵢ + q·pᵢ₋₁ = 0

Assume solution of form pᵢ = xⁱ (characteristic equation approach).

Substituting: p·x² − x + q = 0 → px² − x + (1−p) = 0
Factor: (x − 1)(px − (1−p)) = 0 → (x−1)(px − q) = 0

Roots: x = 1 and x = q/p

CASE 1: p ≠ q (unfair game)

General solution: pᵢ = A·1ⁱ + B·(q/p)ⁱ = A + B·(q/p)ⁱ
Apply boundary conditions:
• p₀ = 0: A + B = 0 → B = −A
• pₙ = 1: A + B·(q/p)ᴺ = 1 → A[1 − (q/p)ᴺ] = 1 → A = 1/[1 − (q/p)ᴺ]

CASE 2: p = q = 1/2 (fair game)

Repeated root x = 1, so general solution: pᵢ = A + B·i
Apply boundary conditions: p₀=0 → A=0; pₙ=1 → B=1/N
Result: pᵢ = i/N

🔢 Formula: Gambler's Ruin — Final Answer

If p ≠ q: pᵢ = [1 − (q/p)ⁱ] / [1 − (q/p)ᴺ]

If p = q = 1/2: pᵢ = i/N

where: i = starting wealth, N = target wealth, p = P(win each round)

✏️Example: Fair Game: p = q = 1/2

Start with $3, target $10, fair coin (p = 0.5).
P(reach $10) = i/N = 3/10 = 0.30
Interpretation: Even with a fair game, you only win 30% of the time when you're starting poor!

✏️Example: Unfair Game: p = 0.4, q = 0.6

Start with $5, target $10. q/p = 0.6/0.4 = 1.5
Numerator: 1 − (1.5)⁵ = 1 − 7.594 = −6.594
Denominator: 1 − (1.5)¹⁰ = 1 − 57.665 = −56.665
P(reach $10) = −6.594 / −56.665 ≈ 0.1164
 Only ~11.6% chance of winning when the odds are against you!

⚡ Theorem: Key Insight from Gambler's Ruin

If p < 0.5 (unfair game against you), P(ruin) → 1 as N → ∞.
Even a slightly unfair game is catastrophic over time (casino advantage!).
If p > 0.5, there is a positive probability of reaching any finite N.
This explains why casinos always win in the long run.

⚠️Common Mistake
The formula pᵢ = i/N only applies when p = q = 1/2. For any other p, you MUST use the (q/p) formula.
Don't forget: p₀ = 0 and pₙ = 1 are the boundary conditions — without them you can't find A and B.
The recurrence pᵢ = p·pᵢ₊₁ + q·pᵢ₋₁ comes from First Step Analysis, not intuition.
 LECTURE 7 — PART 2: RANDOM VARIABLES (INTRODUCTION)
Random variables are one of the most fundamental concepts in probability. They allow us to assign NUMBERS to
outcomes, making calculations much more powerful.

What is a Random Variable?

📘 Definition: Random Variable (R.V.)
A random variable X is a FUNCTION that maps outcomes of a sample space S to real numbers.
Formally: X : S → ℝ
Key point: A random variable is NOT random, and it is NOT a variable.
It is a deterministic FUNCTION — but its INPUT (the outcome) is random.

✏️Example: Coin Toss Random Variable

Sample space: S = {H, T}
Define X = number of heads. Then: X(H) = 1, X(T) = 0
X is a function from {H,T} to {0,1}.
We write: P(X = 1) = P({H}) = 0.5 and P(X = 0) = P({T}) = 0.5

✏️Example: Two Dice Random Variable

S = all pairs (i,j) with i,j ∈ {1,...,6}. |S| = 36
Define X = sum of two dice. Then X maps each pair to a number from 2 to 12.
E.g., X(3,4) = 7, X(1,1) = 2, X(6,6) = 12
P(X = 7) = 6/36 = 1/6 (there are 6 pairs that sum to 7)

Types of Random Variables

• Discrete R.V.: takes a countable (finite or countably infinite) set of values. E.g., counts, integers.
• Continuous R.V.: takes values in an interval of real numbers. E.g., height, time, temperature.

💡 Note: Why Random Variables Matter

R.V.s let us work with numbers instead of abstract outcomes.
We can compute probabilities, averages, variances — all using algebra.
Different random variables can share the same distribution (same 'pattern of randomness').
The distribution of X completely describes the probabilistic behavior of X.

LECTURE 7 — PART 3: BERNOULLI DISTRIBUTION

📘 Definition: Bernoulli Distribution [X ~ Bern(p)]
X is Bernoulli(p) if X takes only two values: 1 (success) and 0 (failure).
P(X = 1) = p (probability of success)
P(X = 0) = 1 − p = q (probability of failure)
Parameter: p ∈ [0, 1]
 Think of it as: one trial of a coin flip, where Heads = 1, Tails = 0.

🔢 Formula: Bernoulli PMF, Mean, Variance

PMF: P(X = k) = pᵏ (1−p)¹⁻ᵏ for k ∈ {0, 1}

Mean: E(X) = p
Variance: Var(X) = p(1−p) = pq

📐 Derivation: Mean of Bernoulli

E(X) = 0 · P(X=0) + 1 · P(X=1)
= 0 · (1−p) + 1 · p
=p ✓
📐 Derivation: Variance of Bernoulli
E(X²) = 0² · (1−p) + 1² · p = p
Var(X) = E(X²) − [E(X)]² = p − p² = p(1−p) = pq ✓

✏️Example: Biased Coin

A coin lands Heads with probability 0.7. Let X = 1 if Heads, 0 if Tails.
X ~ Bern(0.7)
P(X=1) = 0.7, P(X=0) = 0.3
E(X) = 0.7 (average outcome over many flips)
Var(X) = 0.7 × 0.3 = 0.21
💡 Note: Indicator Random Variable
A Bernoulli(p) R.V. is also called an INDICATOR random variable for event A.
We write Iₐ = 1 if event A occurs, 0 if not.
P(Iₐ = 1) = P(A) = p
This is fundamental to the 'Fundamental Bridge': E(Iₐ) = P(A). (Lecture 9)
 LECTURE 7 — PART 4: BINOMIAL DISTRIBUTION
Building the Binomial from Bernoulli
The Binomial is simply the SUM of n independent Bernoulli trials. This is how it naturally arises.
📘 Definition: Binomial Distribution [X ~ Bin(n, p)]
X ~ Bin(n, p) if X = number of successes in n independent Bernoulli(p) trials.
n = number of trials (positive integer)
p = probability of success on each trial (0 ≤ p ≤ 1)
X can take values: 0, 1, 2, 3, ..., n
Think of it as: flip a biased coin n times, count the number of Heads.

PMF Derivation — Where does C(n,k) come from?

📐 Derivation: Binomial PMF Derivation
Question: P(X = k) = P(exactly k successes in n trials)?

Step 1: How many ways can exactly k successes occur in n trials?

→ Choose which k of the n trials are successes: C(n,k) ways

Step 2: What is the probability of ONE specific arrangement with k successes?

→ Each success has prob p, each failure has prob (1−p)
→ P(specific arrangement) = pᵏ · (1−p)ⁿ⁻ᵏ

Step 3: Multiply (each arrangement is disjoint):

P(X = k) = C(n,k) · pᵏ · (1−p)ⁿ⁻ᵏ

🔢 Formula: Binomial PMF

P(X = k) = C(n,k) · pᵏ · (1−p)ⁿ⁻ᵏ for k = 0, 1, 2, ..., n

where C(n,k) = n! / (k!(n−k)!)

🔢 Formula: Binomial Mean and Variance
E(X) = np
Var(X) = np(1−p) = npq

(Derivation will be done formally in Lecture 9 using linearity of expectation.)

⚡ Theorem: Validity of Binomial PMF (Sum = 1)

We must verify: Σ P(X=k) for k=0 to n equals 1.
Σ C(n,k)·pᵏ·qⁿ⁻ᵏ = (p + q)ⁿ [by Binomial Theorem]
= (p + (1−p))ⁿ = 1ⁿ = 1 ✓
The Binomial Theorem guarantees the PMF is valid!
✏️Example: Quality Control
A factory produces items, each defective with probability 0.05.
A sample of n=20 items is inspected. Let X = number of defective items.
X ~ Bin(20, 0.05)

P(X = 0) = C(20,0)·(0.05)⁰·(0.95)²⁰ = 1·1·(0.95)²⁰ ≈ 0.3585

P(X = 1) = C(20,1)·(0.05)¹·(0.95)¹⁹ = 20·0.05·(0.95)¹⁹ ≈ 0.3774
P(X = 2) = C(20,2)·(0.05)²·(0.95)¹⁸ = 190·0.0025·(0.95)¹⁸ ≈ 0.1887

E(X) = 20 × 0.05 = 1 (expect 1 defective item on average)

Var(X) = 20 × 0.05 × 0.95 = 0.95

✏️Example: Coin Toss: P(at least 3 Heads in 5 flips)

X ~ Bin(5, 0.5). Find P(X ≥ 3).

P(X=3) = C(5,3)·(0.5)³·(0.5)² = 10·0.03125 = 0.3125

P(X=4) = C(5,4)·(0.5)⁴·(0.5)¹ = 5·0.03125 = 0.15625
P(X=5) = C(5,5)·(0.5)⁵·(0.5)⁰ = 1·0.03125 = 0.03125

P(X ≥ 3) = 0.3125 + 0.15625 + 0.03125 = 0.5

(Makes sense by symmetry — equally likely to get ≥3 or ≤2 heads!)

⚠️Common Mistake
In Binomial, trials must be INDEPENDENT. If sampling without replacement from a finite population, use
Hypergeometric instead.
P(X = k) requires k to be an integer from 0 to n. P(X = −1) or P(X = n+1) = 0.
Don't confuse n (number of trials) with N (population size in Hypergeometric).
LECTURE 8 — PART 1: PMF AND CDF
Probability Mass Function (PMF)
📘 Definition: Probability Mass Function (PMF)
For a discrete R.V. X, the PMF is the function p(x) = P(X = x).
It gives the probability that X takes EXACTLY the value x.
Valid PMF requirements:
(1) p(x) ≥ 0 for all x
(2) Σ p(x) = 1 (sum over all possible values of x)

✏️Example: PMF of a Fair Die

X = outcome of rolling a fair 6-sided die.
PMF: P(X = k) = 1/6 for k = 1, 2, 3, 4, 5, 6
Verify: Σ P(X=k) = 6 × (1/6) = 1 ✓

Cumulative Distribution Function (CDF)

📘 Definition: Cumulative Distribution Function (CDF)
For a R.V. X, the CDF is F(x) = P(X ≤ x).
It gives the probability that X is at most x.
The CDF is defined for ALL real x (not just integers).
CDF applies to BOTH discrete and continuous random variables.

⚡ Theorem: Properties of CDF — Must Memorize!

1. Non-decreasing: if x₁ < x₂, then F(x₁) ≤ F(x₂)
2. Right-continuous: lim(t→x⁺) F(t) = F(x) for all x
3. Limits: lim(x→−∞) F(x) = 0 and lim(x→+∞) F(x) = 1
4. Jump size at x equals PMF: P(X = x) = F(x) − F(x⁻)
where F(x⁻) = lim(t→x⁻) F(t) (left-hand limit)

🔢 Formula: CDF from PMF (Discrete Case)

F(x) = P(X ≤ x) = Σ P(X = k) for all k ≤ x

Conversely: P(X = k) = F(k) − F(k−1) for integers k

Also useful: P(a < X ≤ b) = F(b) − F(a)

P(X > x) = 1 − F(x)

✏️Example: CDF of a Fair Die

X = outcome of rolling a fair 6-sided die. PMF: P(X=k) = 1/6.
 F(x) = 0 for x < 1
F(x) = 1/6 for 1 ≤ x < 2
F(x) = 2/6 = 1/3 for 2 ≤ x < 3
F(x) = 3/6 = 1/2 for 3 ≤ x < 4
F(x) = 4/6 = 2/3 for 4 ≤ x < 5
F(x) = 5/6 for 5 ≤ x < 6
F(x) = 1 for x ≥ 6

The CDF is a STAIRCASE function, jumping by 1/6 at each integer 1–6.

P(2 < X ≤ 5) = F(5) − F(2) = 5/6 − 2/6 = 3/6 = 1/2 ✓

✏️Example: Using CDF for Binomial

X ~ Bin(5, 0.5). Find P(1 < X ≤ 4).

P(1 < X ≤ 4) = F(4) − F(1)

F(4) = P(X≤4) = P(X=0)+P(X=1)+P(X=2)+P(X=3)+P(X=4)
= 1/32 + 5/32 + 10/32 + 10/32 + 5/32 = 31/32
F(1) = P(X≤1) = 1/32 + 5/32 = 6/32 = 3/16

P(1 < X ≤ 4) = 31/32 − 6/32 = 25/32 ≈ 0.781

⚠️Common Mistake
The CDF F(x) = P(X ≤ x) uses ≤, NOT <. For discrete R.V.s, this matters!
P(X < k) = F(k−1) ≠ F(k) for integer k.
The CDF is RIGHT-continuous, meaning it includes the jump at each value.
Never confuse P(X = x) with F(x). They are completely different!
 LECTURE 8 — PART 2: HYPERGEOMETRIC DISTRIBUTION
The Hypergeometric distribution answers: 'How many successes when sampling WITHOUT replacement?'. It is the
'honest Binomial' when the population is finite.

Setup and Definition

📘 Definition: Hypergeometric Distribution [X ~ HGeom(w, b, n)]
Population: N = w + b objects total.
w = number of 'white' objects (successes)
b = number of 'black' objects (failures)
Draw n objects WITHOUT replacement.
X = number of white objects in the drawn sample.
X ~ HGeom(w, b, n)
X can take values: max(0, n−b) ≤ X ≤ min(w, n)

📐 Derivation: Hypergeometric PMF Derivation

Question: P(X = k) for 0 ≤ k ≤ min(w, n)?

Total ways to choose n objects from N=w+b: C(w+b, n) [denominator]

Favorable: choose k white from w whites AND (n−k) black from b blacks:
C(w, k) × C(b, n−k) [numerator]

Therefore: P(X = k) = C(w,k) · C(b, n−k) / C(w+b, n)

🔢 Formula: Hypergeometric PMF

P(X = k) = C(w, k) · C(b, n−k) / C(w+b, n)

E(X) = n · w / (w+b) = n · w/N

(Expectation will be derived in Lecture 9 using linearity.)

✏️Example: Urn Problem

An urn contains 6 red and 4 blue balls. Draw 5 without replacement.
w = 6 (red), b = 4 (blue), n = 5
X = number of red balls drawn. X ~ HGeom(6, 4, 5)

P(X = 3) = C(6,3)·C(4,2) / C(10,5)

= 20 × 6 / 252
= 120/252 ≈ 0.476

E(X) = 5 × 6/10 = 3 (expect 3 red balls in a sample of 5 from 60% red)

✏️Example: Quality Inspection (Without Replacement)
A batch of 20 items has 4 defectives. Inspect 6 items without replacement.
w = 4 (defective), b = 16 (good), n = 6
X = number of defectives found. X ~ HGeom(4, 16, 6)

P(X = 0) = C(4,0)·C(16,6)/C(20,6) = 1·8008/38760 ≈ 0.207

P(X = 1) = C(4,1)·C(16,5)/C(20,6) = 4·4368/38760 ≈ 0.451
P(X = 2) = C(4,2)·C(16,4)/C(20,6) = 6·1820/38760 ≈ 0.281

E(X) = 6 × 4/20 = 1.2 defectives expected

⚡ Theorem: Hypergeometric vs Binomial

Binomial: sampling WITH replacement (or infinite population). Trials are independent.
Hypergeometric: sampling WITHOUT replacement (finite population). Trials are dependent.

When n is small relative to N (population), Hypergeometric ≈ Binomial.

Rule of thumb: if n < 5% of N, use Binomial as an approximation.

Both have the same MEAN = np (where p = w/N for Hypergeometric).

But Hypergeometric has SMALLER variance (removing an item 'reduces uncertainty').

⚠️Common Mistake
Hypergeometric requires sampling WITHOUT replacement. If with replacement → use Binomial.
The range of X is not always 0 to n. Check: X ≥ max(0, n−b) and X ≤ min(w, n).
In the PMF, if k > w or n−k > b, then P(X=k) = 0 (impossible).
LECTURE 8 — PART 3: INDEPENDENCE OF RANDOM VARIABLES
📘 Definition: Independence of Random Variables
Random variables X and Y are INDEPENDENT if for all values a, b:
P(X = a, Y = b) = P(X = a) · P(Y = b)

Equivalently: knowing the value of X gives NO information about Y.

Notation: X ⊥ Y

⚡ Theorem: Key Property: Sum of Independent Binomials

If X ~ Bin(n, p) and Y ~ Bin(m, p), and X ⊥ Y, then:
X + Y ~ Bin(n+m, p)

This makes sense: n trials + m trials = n+m trials, same success probability.
CRITICAL: The p must be the SAME for both!

📐 Derivation: Proof: Sum of Independent Binomials

P(X+Y = k) = Σⱼ P(X=j)·P(Y=k−j) [by independence and convolution]

= Σⱼ C(n,j)·pʲ·qⁿ⁻ʲ · C(m,k−j)·pᵏ⁻ʲ·qᵐ⁻⁽ᵏ⁻ʲ⁾

= pᵏ·qⁿ⁺ᵐ⁻ᵏ · Σⱼ C(n,j)·C(m,k−j)

By Vandermonde's identity: Σⱼ C(n,j)·C(m,k−j) = C(n+m, k)

Therefore: P(X+Y=k) = C(n+m, k)·pᵏ·qⁿ⁺ᵐ⁻ᵏ = Bin(n+m, p) ✓

✏️Example: Basketball Free Throws

Player A makes free throws with p=0.7. Takes 10 shots → X ~ Bin(10, 0.7).
Player B also has p=0.7. Takes 8 shots → Y ~ Bin(8, 0.7). X ⊥ Y.
Total shots made: X + Y ~ Bin(18, 0.7).
E(X+Y) = 18 × 0.7 = 12.6 shots made on average.
EXAM QUESTIONS & SOLUTIONS (Lectures 7 & 8)

❓ Exam Q: Gambler starts with $4, target $8, fair game. Find P(ruin).
Answer: P(reach $8) = 4/8 = 1/2. So P(ruin) = 1 − 1/2 = 1/2.

❓ Exam Q: Gambler starts with $3, target $9, p=0.4, q=0.6. Find P(winning).
Answer: q/p = 1.5. P(win) = [1−(1.5)³]/[1−(1.5)⁹] = [1−3.375]/[1−38.44] = (−2.375)/(−37.44) ≈ 0.0635
(about 6.4%)

❓ Exam Q: X ~ Bin(10, 0.3). Find P(X = 4) and E(X).

Answer: P(X=4) = C(10,4)·(0.3)⁴·(0.7)⁶ = 210·0.0081·0.1176 ≈ 0.2001 E(X) = 10×0.3 = 3

❓ Exam Q: State the three properties that a valid CDF must satisfy.
Answer: (1) Non-decreasing. (2) Right-continuous. (3) F(−∞)=0 and F(+∞)=1.

❓ Exam Q: Bowl has 7 green and 5 yellow marbles. Draw 4 without replacement. P(exactly 2 green)?
Answer: X ~ HGeom(7, 5, 4). P(X=2) = C(7,2)·C(5,2)/C(12,4) = 21×10/495 = 210/495 ≈ 0.424

❓ Exam Q: Why can't we use Binomial for sampling WITHOUT replacement from a finite population?
Answer: Binomial requires INDEPENDENT trials. Without replacement, the outcome of each draw affects
the next (the pool changes), so trials are dependent. Use Hypergeometric instead.

❓ Exam Q: X ~ Bern(0.6). Find E(X), Var(X), and P(X=1).

Answer: P(X=1) = 0.6, E(X) = 0.6, Var(X) = 0.6×0.4 = 0.24

❓ Exam Q: X ~ Bin(n,p) and Y ~ Bin(m,p) are independent. What is the distribution of X+Y?
Answer: X+Y ~ Bin(n+m, p) (same success probability p must hold for both)
COMMON MISTAKES — LECTURES 7 & 8

⚠️Common Mistake
GAMBLER'S RUIN: Using pᵢ = i/N when p ≠ 0.5. This formula is ONLY for the fair game!

⚠️Common Mistake
RANDOM VARIABLES: Thinking a R.V. is 'random'. It is a deterministic FUNCTION. The randomness comes
from the random experiment, not from X itself.

⚠️Common Mistake
BINOMIAL vs HYPERGEOMETRIC: Using Binomial when sampling without replacement from a finite
population. Without replacement → trials are dependent → Hypergeometric!

⚠️Common Mistake
CDF: Writing P(X ≤ x) = Σ P(X=k) for k < x instead of k ≤ x. The CDF includes x!

⚠️Common Mistake
PMF: Forgetting to include C(n,k) in the Binomial PMF. The formula is C(n,k)·pᵏ·qⁿ⁻ᵏ, not just pᵏ·qⁿ⁻ᵏ.

⚠️Common Mistake
SUM OF BINOMIALS: Adding Bin(n,p) + Bin(m,q) where p≠q and claiming the result is Binomial. Only
works if p = q!

⚠️Common Mistake
HYPERGEOMETRIC RANGE: Assuming X goes from 0 to n. The true range is max(0, n−b) to min(w, n).
 ⚡ QUICK REFERENCE SUMMARY SHEET

Distribution Comparison Table

Distribution Parameters PMF Mean Variance

Bernoulli Bern(p) p ∈ [0,1] pᵏ(1−p)¹⁻ᵏ k∈{0,1} p p(1−p)

n≥1, p∈[0,1] C(n,k)pᵏ(1−p)ⁿ⁻ᵏ np np(1−p)

k=0..n
Binomial Bin(n,p)

w,b≥0, n≥1 C(w,k)C(b,n−k) n·w/(w+b)

────────────
Hypergeometric see Lec 9

C(w+b, n)
HGeom(w,b,n)

All Key Formulas

Concept Formula
Gambler's Ruin (p ≠ q) pᵢ = [1 − (q/p)ⁱ] / [1 − (q/p)ᴺ]

Gambler's Ruin (p = q = ½) pᵢ = i/N

Bernoulli PMF P(X=k) = pᵏ(1−p)¹⁻ᵏ for k ∈ {0,1}

Binomial PMF P(X=k) = C(n,k)·pᵏ·(1−p)ⁿ⁻ᵏ

Hypergeometric PMF P(X=k) = C(w,k)·C(b,n−k) / C(w+b,n)

CDF definition F(x) = P(X ≤ x)

Prob from CDF P(a < X ≤ b) = F(b) − F(a)

PMF from CDF (discrete) P(X=k) = F(k) − F(k−1)

Sum of Bin R.V.s Bin(n,p) + Bin(m,p) = Bin(n+m,p) [indep.]

Bernoulli Mean/Var E=p, Var=p(1−p)

Binomial Mean/Var E=np, Var=np(1−p)

Statistics 110 Lectures 7 & 8 | Harvard University | Prof. Joe Blitzstein | [Link]/PNrqCdslGi4 | [Link]/k2BB0p8byGA