8.2 Probability Distributions: Binomial & Normal Models

8.2 Probability Distributions: Binomial & Normal Models

Probability distributions bridge theoretical chance and applied statistical inference by assigning probabilities to numerical values of random variables. For secondary mathematics educators, mastering discrete and continuous models is essential for guiding students from counting permutations and combinations to analyzing real-world sampling distributions. Understanding the structural properties of Binomial and Normal distributions—along with their asymptotic connection via the Central Limit Theorem—allows teachers to address common misconceptions regarding discrete endpoints, continuity corrections, and probability densities.

---

1. Discrete Random Variables, PMFs, Expectation & Variance

A random variable $X$ is a measurable function mapping elements of a sample space $S$ to the real numbers: $X: S \to \mathbb{R}$. A random variable is discrete if its range (support) is finite or countably infinite.

Probability Mass Function (PMF)

The distribution of a discrete random variable is completely specified by its probability mass function $p(x) = P(X = x)$, which must satisfy two foundational properties:

1. $p(x) \ge 0$ for all $x \in \mathbb{R}$.
2. $\sum_{x} p(x) = 1$, where the summation extends over all possible values of $X$.

Expected Value (Mean) of a Discrete Random Variable

The expected value (or mathematical expectation) $E[X]$, denoted $\mu$, is the probability-weighted average of all possible values:

$$E[X] = \mu = \sum_{x} x \cdot p(x)$$

Properties of Expectation (Linearity): For any constants $a, b \in \mathbb{R}$ and random variables $X, Y$:

• $E[aX + b] = a E[X] + b$
• $E[X + Y] = E[X] + E[Y]$ (holds universally, even if $X$ and Y are dependent).

Variance and Standard Deviation

The variance $\operatorname{Var}(X)$, denoted $\sigma^2$, measures the expected squared dispersion around the mean:

$$\operatorname{Var}(X) = \sigma^2 = E[(X - \mu)^2] = \sum_{x} (x - \mu)^2 \cdot p(x)$$

Expanding the squared binomial yields the standard computational shortcut:

$$\operatorname{Var}(X) = E[X^2 - 2\mu X + \mu^2] = E[X^2] - 2\mu E[X] + \mu^2 = E[X^2] - (E[X])^2$$

where $E[X^2] = \sum_{x} x^2 \cdot p(x)$. The standard deviation is $\sigma = \sqrt{\operatorname{Var}(X)}$.

Variance Scaling Property: For constants $a, b$:

$$\operatorname{Var}(aX + b) = a^2 \operatorname{Var}(X) \implies \sigma_{aX+b} = |a| \sigma_X$$

---

2. The Binomial Distribution (BINS Framework)

The Binomial distribution models the number of successes in a sequence of independent Bernoulli trials.

The Four BINS Criteria

To establish that a random variable $X$ follows a Binomial distribution, $X \sim B(n, p)$, educators and students verify the BINS acronym:

1. B — Binary Outcomes: Each trial results in exactly one of two mutually exclusive categories: Success ($S$) or Failure ($F$).
2. I — Independent Trials: The outcome of any individual trial does not alter the probabilities of subsequent trials.
3. N — Number of Trials Fixed: The total number of trials $n$ is fixed in advance.
4. S — Same Probability of Success: The probability of success $p = P(S)$ is constant on every single trial ($q = 1 - p = P(F)$).

The Binomial Probability Formula

For $X \sim B(n, p)$, the probability of obtaining exactly $k$ successes in $n$ trials (where $k \in \{0, 1, 2, \dots, n\}$) is:

$$P(X = k) = \binom{n}{k} p^k (1 - p)^{n - k} = \frac{n!}{k!(n - k)!} p^k q^{n - k}$$

Combinatorial Derivation: Any specific ordered sequence containing $k$ successes and $n - k$ failures (e.g., $SS \dots SFF \dots F$) has probability $p^k (1 - p)^{n - k}$ by independence. The binomial coefficient $\binom{n}{k}$ counts the number of distinct ways to choose which $k$ of the $n$ trial slots contain the successes.

Mean and Variance of the Binomial Distribution

Using indicator variables, express $X = I_1 + I_2 + \dots + I_n$, where each $I_j \sim \text{Bernoulli}(p)$ with $P(I_j = 1) = p$ and $P(I_j = 0) = 1 - p$. For each indicator:

• $E[I_j] = 1(p) + 0(1 - p) = p$
• $E[I_j^2] = 1^2(p) + 0^2(1 - p) = p \implies \operatorname{Var}(I_j) = p - p^2 = p(1 - p)$

By linearity of expectation and independence:

$$\mu = E[X] = \sum_{j=1}^n E[I_j] = n p$$ $$\sigma^2 = \operatorname{Var}(X) = \sum_{j=1}^n \operatorname{Var}(I_j) = n p (1 - p) \implies \sigma = \sqrt{n p (1 - p)}$$

---

3. Continuous Probability Distributions & PDFs

A random variable $X$ is continuous if its possible values form an uncountable interval of real numbers (e.g., time, height, mass). Because an uncountable set cannot be summed with a discrete PMF, continuous variables are governed by a probability density function (PDF) $f(x)$.

Axiomatic Properties of a PDF

1. Non-negativity: $f(x) \ge 0$ for all $x \in \mathbb{R}$.
2. Total Area: $\int_{-\infty}^{\infty} f(x) \, dx = 1$.

+-----------------------------------------------------------------------------+
|                   CONTINUOUS PROBABILITY AS AREA UNDER A PDF                |
|                                                                             |
|             f(x) ^                                                          |
|                  |                ****                                      |
|                  |             *        *                                   |
|                  |           *   Area =  *                                  |
|                  |          *  P(a ≤ X ≤ b)*                                |
|                  |         * |////////////| *                               |
|                  +---------+--+----------+--+----------> x                  |
|                              a            b                                 |
|                                                                             |
|       P(a ≤ X ≤ b) = ∫_a^b f(x) dx         P(X = c) = ∫_c^c f(x) dx = 0     |
+-----------------------------------------------------------------------------+

The Point-Probability Zero Paradox

For any continuous random variable $X$ and any specific exact real constant $c$:

$$P(X = c) = \int_c^c f(x) \, dx = 0$$

Instructional Consequence: For continuous distributions, strict and non-strict inequalities are mathematically identical:

$$P(a \le X \le b) = P(a < X \le b) = P(a \le X < b) = P(a < X < b)$$

This contrasts starkly with discrete distributions, where $P(X \le 5) = P(X < 5) + P(X = 5)$.

---

4. The Normal Distribution & Standard Normal Z-Scores

The Normal (Gaussian) distribution $N(\mu, \sigma)$ is the foundational continuous distribution of mathematical statistics, with PDF:

$$f(x) = \frac{1}{\sigma \sqrt{2\pi}} e^{-\frac{(x - \mu)^2}{2\sigma^2}}, \quad x \in \mathbb{R}$$

Structural Characteristics

• Symmetry: The curve is symmetric, bell-shaped, and unimodal about the center $x = \mu$. Mean = Median = Mode = $\mu$.
• Inflection Points: The concavity of $f(x)$ transitions from concave down to concave up exactly at the inflection coordinates $x = \mu - \sigma$ and $x = \mu + \sigma$.
• Asymptotes: The tails approach the horizontal axis asymptotically ($f(x) \to 0$ as $x \to \pm\infty$) but never touch it.

The Empirical Rule (68–95–99.7 Rule)

For any normally distributed population:

• $\approx 68.27\%$ of the distribution lies within 1 standard deviation of the mean: $(\mu - \sigma, \mu + \sigma)$.
• $\approx 95.45\%$ of the distribution lies within 2 standard deviations of the mean: $(\mu - 2\sigma, \mu + 2\sigma)$.
• $\approx 99.73\%$ of the distribution lies within 3 standard deviations of the mean: $(\mu - 3\sigma, \mu + 3\sigma)$.

Standard Normal Distribution $Z \sim N(0, 1)$ & Z-Scores

To evaluate probabilities for an arbitrary $X \sim N(\mu, \sigma)$, the variable is standardized into the standard normal score (z-score):

$$Z = \frac{X - \mu}{\sigma} \sim N(0, 1)$$

The z-score measures the signed distance between an observation $x$ and the population mean $\mu$ in units of standard deviation. The cumulative distribution function (CDF) is denoted $\Phi(z) = P(Z \le z) = \frac{1}{\sqrt{2\pi}} \int_{-\infty}^z e^{-t^2/2} \, dt$.

---

5. Normal Approximation to the Binomial & Continuity Correction

When $n$ is large, evaluating binomial cumulative sums $\sum_{k=a}^b \binom{n}{k} p^k (1 - p)^{n - k}$ by hand or basic calculators is computationally prohibitive. By the De Moivre-Laplace Limit Theorem (a precursor to the Central Limit Theorem), the Binomial distribution converges asymptotically to a Normal distribution with identical mean and variance: $N(np, \sqrt{np(1-p)})$.

Rule of Thumb Conditions

The normal approximation is valid when both expected successes and expected failures are at least 10:

$$n p \ge 10 \quad \text{and} \quad n(1 - p) \ge 10$$

The Continuity Correction Protocol

Because we are approximating a discrete integer-valued random variable $X$ with a continuous density curve $Y$, each integer $k$ on the discrete number line corresponds to a unit-width interval $[k - 0.5, k + 0.5]$ under the continuous curve.

---

6. Secondary Pedagogical Traps & Instructional Interventions

Accomplished mathematics educators anticipate predictable student errors in distribution theory:

Misconception 1: Interchanging Strict and Non-Strict Inequalities Indiscriminately

Students frequently assume $P(X < 4) = P(X \le 4)$ in all probability problems, confusing continuous and discrete rules.

• Pedagogical Intervention: Draw a discrete probability histogram where each integer is represented by a distinct rectangular bar. Show that $P(X \le 4)$ includes the entire bar over $4$, whereas $P(X < 4)$ stops at the bar over $3$. Thus, $P(X \le 4) - P(X < 4) = P(X = 4) > 0$. Then contrast this with a smooth continuous curve where the area over a single point $x = 4$ is a line segment of zero width (area $= 0$).

Misconception 2: Arbitrary Sign Selection in Continuity Corrections

Students memorize that continuity corrections involve $\pm 0.5$, but guess whether to add or subtract.

• Pedagogical Intervention: Mandate that students sketch a segment of the discrete integer axis and shade the "target bars" before writing any numbers. For example, if asked for $P(X \ge 12)$, have them shade the bars $12, 13, 14, \dots$. Visually, the shaded region starts at the left edge of bar $12$, which is $11.5$. Thus, $P(X \ge 12) \approx P(Y \ge 11.5)$. If asked for $P(X > 12)$, the shaded region begins at bar $13$, whose left edge is $12.5$. The visual diagram prevents sign errors.

Misconception 3: Interpreting the PDF Height $f(x)$ as a Probability

When evaluating a normal density function, students often compute $f(0) = \frac{1}{\sqrt{2\pi}} \approx 0.3989$ and conclude that $P(Z = 0) = 39.89\%$.

• Pedagogical Intervention: Remind students that $f(x)$ represents probability density (probability per unit of $x$), not probability itself. Illustrate with a uniform distribution on $[0, 0.5]$ where $f(x) = 2.0$. If $f(x)$ were a probability, $P = 2.0 > 1.0$ would violate Kolmogorov's first and second axioms. Reiterate: only areas under $f(x)$ represent probabilities.