Statistics • Score 5 Strategy

Binomial vs. Geometric Probability Models & Expected Values Guide: AP Statistics Score 5 for Carnegie Mellon University

AP Statistics Mastery Guide: Binomial vs. Geometric Probability Models & Expected Values

1. Introduction & AP Exam Weight

In AP Statistics, discrete probability distributions form the rigorous mathematical foundation for statistical inference. Specifically, Binomial and Geometric Probability Models constitute the backbone of Unit 4: Random Variables, Sets, and Probability Distributions, which directly accounts for 10%–20% of the multiple-choice questions (MCQs) and serves as a frequent centerpiece for Free-Response Questions (FRQs), particularly FRQ #2, #3, or the cumulative FRQ #6 (Investigative Task).

Conceptual Scope

Binomial Models ($\text{Binom}(n, p)$): Count the number of successes $X$ in a fixed number of independent Bernoulli trials $n$.
Geometric Models ($\text{Geom}(p)$): Count the number of independent Bernoulli trials $Y$ required to observe the first success (variable trial count, unbounded sample space).
Expected Value ($\mu$) and Variance ($\sigma^2$): Quantify the central tendency and dispersion of these discrete random variables.

For students targeting a score of 5 and planning to leverage this credit at top-tier institutions like Carnegie Mellon University (CMU), memorizing formulas is insufficient. You must master formal mathematical proofs of expectation, construct algorithmic simulations of these processes, and articulate condition checks with bulletproof precision under AP grading standards.

2. Deep Concept Breakdown

A. Strict Structural Conditions: BINS vs. BITS

Before applying any distribution formula on an AP FRQ, you must explicitly check and state the model assumptions in the context of the problem.

Condition	Binomial Model ($\text{Binom}(n, p)$)	Geometric Model ($\text{Geom}(p)$)
Binary	Outcomes are binary: Success ($S$) or Failure ($F$).	Outcomes are binary: Success ($S$) or Failure ($F$).
Independent	Trials are independent (or $n \le 0.10N$ for finite sampling without replacement).	Trials are independent.
Number / Trials	Fixed number of trials $n$ defined prior to experiment.	Trials continue until First success observed.
Same $p$	Probability of success $p$ is constant across all trials.	Probability of success $p$ is constant across all trials.
Mnemonic	B - I - N - S	B - I - T - S

B. Binomial Distribution Model $X \sim \text{Binom}(n, p)$

Probability Mass Function (PMF): $$P(X = k) = \binom{n}{k} p^k (1-p)^{n-k}, \quad \text{for } k \in {0, 1, 2, \dots, n}$$ where $\binom{n}{k} = \frac{n!}{k!(n-k)!}$.
Cumulative Distribution Function (CDF): $$P(X \le k) = \sum_{j=0}^{k} \binom{n}{j} p^j (1-p)^{n-j}$$

Rigorous Proof of Expected Value $E[X] = np$

Using the definition of expectation for discrete random variables:

$$E[X] = \sum_{k=0}^{n} k \cdot P(X = k) = \sum_{k=0}^{n} k \cdot \frac{n!}{k!(n-k)!} p^k (1-p)^{n-k}$$

Since the $k=0$ term equals zero:

$$E[X] = \sum_{k=1}^{n} \frac{n!}{(k-1)!(n-k)!} p^k (1-p)^{n-k}$$

Factor out $n$ and $p$:

$$E[X] = np \sum_{k=1}^{n} \frac{(n-1)!}{(k-1)!((n-1)-(k-1))!} p^{k-1} (1-p)^{(n-1)-(k-1)}$$

Let $j = k - 1$ and $m = n - 1$. As $k$ ranges from $1$ to $n$, $j$ ranges from $0$ to $m$:

$$E[X] = np \sum_{j=0}^{m} \frac{m!}{j!(m-j)!} p^j (1-p)^{m-j} = np \sum_{j=0}^{m} \binom{m}{j} p^j (1-p)^{m-j}$$

By the Binomial Theorem, $\sum_{j=0}^{m} \binom{m}{j} p^j (1-p)^{m-j} = (p + (1-p))^m = 1^m = 1$.

$$\therefore E[X] = np$$

Variance Derivation Summary: $$\text{Var}(X) = E[X(X-1)] + E[X] - (E[X])^2 = n(n-1)p^2 + np - n^2p^2 = np(1-p)$$ $$\sigma_X = \sqrt{np(1-p)}$$

C. Geometric Distribution Model $Y \sim \text{Geom}(p)$

Probability Mass Function (PMF): $$P(Y = k) = (1-p)^{k-1} p, \quad \text{for } k \in {1, 2, 3, \dots}$$
Tail Probability Function (Crucial for AP FRQs): $$P(Y > k) = (1-p)^k$$ Proof: To require more than $k$ trials for the first success, the first $k$ trials must all be failures.
Cumulative Distribution Function (CDF): $$P(Y \le k) = 1 - P(Y > k) = 1 - (1-p)^k$$

Rigorous Proof of Expected Value $E[Y] = \frac{1}{p}$ (Recursive Expectation Method)

By total expectation law, conditioning on the outcome of the first trial:

$$E[Y] = P(\text{Success on Trial 1}) \cdot 1 + P(\text{Failure on Trial 1}) \cdot (1 + E[Y])$$

Substitute $P(\text{Success}) = p$ and $P(\text{Failure}) = 1-p$:

$$E[Y] = p(1) + (1-p)(1 + E[Y])$$ $$E[Y] = p + 1 - p + (1-p)E[Y]$$ $$E[Y] = 1 + (1-p)E[Y]$$ $$E[Y] - (1-p)E[Y] = 1$$ $$p E[Y] = 1 \implies E[Y] = \frac{1}{p}$$

Variance: $$\text{Var}(Y) = \frac{1-p}{p^2}, \quad \sigma_Y = \frac{\sqrt{1-p}}{p}$$

D. Algorithmic Simulation & Verification in Python

To consolidate intuition for stochastic convergence, the following computational module computes exact PMFs, cumulative distributions, and empirical moments for both probability models.

import numpy as np
import scipy.stats as stats

def analyze_discrete_models(p: float, n_binom: int, num_simulations: int = 1_000_000):
    """
    Simulates and validates Theoretical vs Empirical moments for 
    Binomial and Geometric Probability Models.
    """
    np.random.seed(42)

    # ------------------ BINOMIAL DISTRIBUTION ------------------
    binom_theoretical_mean = n_binom * p
    binom_theoretical_var = n_binom * p * (1 - p)

    binom_simulated = np.random.binomial(n=n_binom, p=p, size=num_simulations)
    binom_empirical_mean = np.mean(binom_simulated)
    binom_empirical_var = np.var(binom_simulated, ddof=1)

    # ------------------ GEOMETRIC DISTRIBUTION ------------------
    geom_theoretical_mean = 1 / p
    geom_theoretical_var = (1 - p) / (p ** 2)

    # numpy.random.geometric counts trials until first success (1-indexed)
    geom_simulated = np.random.geometric(p=p, size=num_simulations)
    geom_empirical_mean = np.mean(geom_simulated)
    geom_empirical_var = np.var(geom_simulated, ddof=1)

    print("=== BINOMIAL MODEL RESULTS ===")
    print(f"Theoretical E[X]: {binom_theoretical_mean:.4f} | Empirical E[X]: {binom_empirical_mean:.4f}")
    print(f"Theoretical Var(X): {binom_theoretical_var:.4f} | Empirical Var(X): {binom_empirical_var:.4f}\n")

    print("=== GEOMETRIC MODEL RESULTS ===")
    print(f"Theoretical E[Y]: {geom_theoretical_mean:.4f} | Empirical E[Y]: {geom_empirical_mean:.4f}")
    print(f"Theoretical Var(Y): {geom_theoretical_var:.4f} | Empirical Var(Y): {geom_empirical_var:.4f}")

if __name__ == "__main__":
    analyze_discrete_models(p=0.25, n_binom=20)

3. Common AP Exam Pitfalls & Score 5 Scoring Rubric Nuances

To secure a Score 5, your FRQ answers must earn "Essentially Correct" (E) ratings across all components. Below are the structural mistakes that demote solutions from a 5 to a 4 (or 3).

Pitfall 1: Calculator Syntax Without Contextual Parameter Definitions

The Score 4 Mistake: Writing binompdf(10, 0.3, 4) = 0.2001 or geometcdf(0.15, 3).
AP Rubric Reality: Calculator syntax alone is penalized. If work is shown as binompdf(10, 0.3, 4), it receives Partial Credit (P) or Incorrect (I).
The Score 5 Solution:
Define the random variable: "Let $X$ = the number of defective units out of $n = 10$."
State the model and parameters explicitly: "$X \sim \text{Binom}(n = 10, p = 0.3)$."
Write the algebraic formula with values substituted: $$P(X = 4) = \binom{10}{4} (0.3)^4 (0.7)^6 \approx 0.2001$$

Pitfall 2: Omitting the Independent Trials / 10% Condition Verification

The Score 4 Mistake: Stating "It is binomial because there are two outcomes and constant probability."
The Score 5 Solution: When sampling without replacement, you must explicitly evaluate the 10% condition: $$\text{"The sample size } n = 50 \text{ is less than } 10\% \text{ of the total population of } N = 2,000 \text{ items } (50 \le 0.10 \times 2000), \text{ ensuring trial independence."}$$

Pitfall 3: Boundary Errors in Geometric Tail Probabilities

The Score 4 Mistake: Calculating $P(Y \ge 5)$ as $(1-p)^5$.
The Score 5 Correction:
"At least 5 trials to get the first success" means the first 4 trials were failures: $$P(Y \ge 5) = P(Y > 4) = (1-p)^4$$
"More than 5 trials to get the first success": $$P(Y > 5) = (1-p)^5$$

4. Carnegie Mellon University Placement Pathway

      AP Statistics Exam (Score 5)
                   │
                   ▼
       Waive: 36-200 (9 Units)
       "Reasoning with Data"
                   │
                   ▼
  Accelerate directly into Core Sequence:
36-225: Introduction to Probability Theory
                   │
                   ▼
   Advanced Machine Learning & AI
 (CS 10-701, ML 10-315, LTI 11-711)

Strategic Academic Placement

At Carnegie Mellon University (CMU), achieving a Score 5 on AP Statistics awards 9 units of credit for 36-200: Reasoning with Data.

Course Exemption Advantage: Exempting 36-200 frees 9 units in your freshman course plan, satisfying a foundational quantitative requirement for the Dietrich College of Humanities and Social Sciences, the School of Computer Science (SCS), and the Tepper School of Business.
Immediate Acceleration Track: Bypassing the introductory survey module allows high-achieving STEM majors to register directly for 36-225: Introduction to Probability Theory, followed immediately by 36-226: Introduction to Statistical Inference.
Core Application to Machine Learning (ML) & Data Science:
Binomial models transition directly into Bernoulli Process Modeling, Binomial Naive Bayes Classifiers, and Logistic Regression loss formulations.
Geometric models serve as the discrete foundational analog for Exponential Waiting-Time Models, Reinforcement Learning Discounted Horizons, and Markov Chain First-Passage Times.

5. High-Yield Practice Problem & Step-by-Step Solution Checklist

Problem Statement

An autonomous robotic quality assurance scanner at a manufacturing plant inspects microchips produced on an assembly line. The production line operates under stable conditions where each inspected microchip has an independent probability $p = 0.08$ of containing a fabrication defect.

Part (a): Calculate the probability that the scanner identifies the first defective microchip on exactly the 6th inspection.
Part (b): Calculate the probability that the scanner requires more than 8 inspections to find the first defective microchip.
Part (c): The plant supervisor checks a batch of $n = 25$ microchips. Let $X$ represent the number of defective microchips found in this batch. Calculate $P(X \ge 3)$.
Part (d): The operation cost of inspecting a batch of 25 microchips is governed by a cost function: $C(X) = 150 + 12X + 3X^2$ dollars, where $X$ is the number of defective microchips. Calculate the expected inspection cost $E[C(X)]$ for a batch of 25 microchips.

Complete Solution & AP Scoring Checklist

Solution to Part (a)

Define Random Variable & Distribution: Let $Y = \text{the number of microchip inspections until the first defect is found}$. Since trials are independent, outcomes are binary (Defective / Non-defective), $p = 0.08$ is constant, and we inspect until the first success, $Y \sim \text{Geom}(p = 0.08)$.
Calculation: $$P(Y = 6) = (1 - 0.08)^{6-1} (0.08) = (0.92)^5 (0.08)$$ $$P(Y = 6) = (0.65908) (0.08) \approx 0.0527$$

Solution to Part (b)

Identify Equivalent Event: "More than 8 inspections required to find the first defect" corresponds to $P(Y > 8)$.
Apply Geometric Tail Probability: This is equivalent to observing 8 consecutive non-defective microchips: $$P(Y > 8) = (1 - p)^8 = (0.92)^8 \approx 0.5132$$

Solution to Part (c)

Define Variable & Validate Model Conditions (BINS): Let $X = \text{number of defective microchips in a batch of } n = 25$.
B: Binary outcomes (Defective vs. Non-Defective).
I: Independent trials (Each microchip defect status is independent).
N: Fixed number of trials ($n = 25$).
S: Same success probability ($p = 0.08$). Therefore, $X \sim \text{Binom}(n = 25, p = 0.08)$.
Calculation using Complement Rule: $$P(X \ge 3) = 1 - P(X \le 2) = 1 - [P(X=0) + P(X=1) + P(X=2)]$$ $$P(X = 0) = \binom{25}{0} (0.08)^0 (0.92)^{25} \approx 0.1244$$ $$P(X = 1) = \binom{25}{1} (0.08)^1 (0.92)^{24} = 25(0.08)(0.1352) \approx 0.2704$$ $$P(X = 2) = \binom{25}{2} (0.08)^2 (0.92)^{23} = 300(0.0064)(0.1470) \approx 0.2822$$ $$P(X \le 2) = 0.1244 + 0.2704 + 0.2822 = 0.6770$$ $$P(X \ge 3) = 1 - 0.6770 = 0.3230$$

Solution to Part (d)

Apply Linearity of Expectation to Function of $X$: $$E[C(X)] = E[150 + 12X + 3X^2] = 150 + 12 E[X] + 3 E[X^2]$$
Calculate First and Second Moments of $X \sim \text{Binom}(25, 0.08)$:
$E[X] = np = 25 \times 0.08 = 2.0$
$\text{Var}(X) = np(1-p) = 25 \times 0.08 \times 0.92 = 1.84$
Utilize Variance Identity for $E[X^2]$: $$\text{Var}(X) = E[X^2] - (E[X])^2 \implies E[X^2] = \text{Var}(X) + (E[X])^2$$ $$E[X^2] = 1.84 + (2.0)^2 = 1.84 + 4.0 = 5.84$$
Compute Expected Cost: $$E[C(X)] = 150 + 12(2.0) + 3(5.84)$$ $$E[C(X)] = 150 + 24 + 17.52 = 191.52$$

Final Answer Statement: The expected inspection cost for a batch of 25 microchips is $191.52.

AP Reader Scoring Guidelines Checklist (FRQ Verification)

Component	Essential Requirements for Score 5 ("Essentially Correct")
Part (a)	Identifies geometric distribution with $p=0.08$; shows explicit work using $(0.92)^5(0.08)$; yields correct value (~0.0527).
Part (b)	Sets up tail probability $P(Y > 8)$ or $1 - P(Y \le 8)$; uses formula $(0.92)^8$; yields correct answer (~0.5132).
Part (c)	Identifies binomial distribution with $n=25, p=0.08$; explicitly shows setup for $P(X \ge 3)$ via binomial cumulative summation or complement; provides correct numeric value (~0.3230).
Part (d)	Uses variance identity $\text{Var}(X) = E[X^2] - (E[X])^2$ to correctly solve $E[X^2] = 5.84$; applies expectation operators algebraically; arrives at final cost $191.52 with context (dollars).

Aiming for a Score 5 in Statistics?

Secure admission and advanced standing at top institutions like Carnegie Mellon University with elite 1-on-1 AP STEM mentorship.

無料相談・学習プラン診断