Chapter 6 Normal Distribution: Theory, Formulas, and Applications

Fundamental Characteristics of the Normal Distribution

• Continuous Probability Distribution: The normal distribution is a continuous distribution and represents the most widely utilized continuous model in theoretical and applied statistics.

• Cross-Disciplinary Applications: The distribution's signature "bell curve" appears ubiquitously across numerous fields, including:

  • Psychology (e.g., standardized cognitive testing and behavioral metrics)

  • Business and Economics (e.g., market forecasting, quality control, asset returns)

  • Natural and Physical Sciences (e.g., biological traits, experimental error analysis)

  • Nursing and Health Sciences (e.g., blood pressure, biometric monitoring)

  • Mathematics and Statistics

• Empirical Demonstration (Shoe Sizes): Gathering shoe sizes from a sufficiently large random sample of individuals yields a bell-shaped histogram that closely matches a theoretical normal curve.

• Distribution Parameters: A normal distribution is completely defined by two numerical descriptive parameters:

  • Mean ($\mu$): The measure of central tendency locating the center of the curve.

  • Standard Deviation ($\sigma$): The measure of spread or dispersion determining curve width.

• Distributional Notation: When a measured continuous random variable $X$ is normally distributed with mean $\mu$ and standard deviation $\sigma$, it is formally expressed as:     $X \sim N(\mu, \sigma)$

• Symmetry and Central Equivalence:

  • The curve is mathematically symmetric about a vertical line positioned at $x = \mu$.

  • Because the curve is symmetrical about the center, the theoretical mean is equal to the median: $\text{Mean} = \text{Median} = \mu$.

• Total Probability Area: The total area bounded under the normal curve and above the horizontal axis is strictly equal to $1$ (or $100\%$):     $\int_{-\infty}^{\infty} f(x)\,dx = 1$

• Influence of Parameters on Graphical Shape:

  • Effect of Changing Standard Deviation ($\sigma$): Changing $\sigma$ alters the dispersion. A smaller $\sigma$ concentrates data around $\mu$, making the curve taller and narrower; a larger $\sigma$ flattens and widens the curve.

  • Effect of Changing Mean ($\mu$): Changing $\mu$ produces a horizontal rigid shift of the distribution along the horizontal axis to the left or right without altering its geometric spread.

  • Infinite Family of Curves: Because $\mu$ and $\sigma$ can take an infinite variety of real values ($\sigma > 0$), there exists an infinite family of distinct normal probability curves.

Mathematical Formulation and Historical Milestones

• Probability Density Function (PDF): The mathematical equation describing the normal distribution curve is expressed as:     $f(x) = \frac{1}{\sigma \sqrt{2\pi}} \cdot e^{-\frac{1}{2}\left(\frac{x - \mu}{\sigma}\right)^2}$

  • where $e \approx 2.71828$

  • where $\pi \approx 3.14159$

  • where $-\infty < x < \infty$

• Chronological Development of the Normal Model:

  • 1733 (Abraham de Moivre): DeMoivre mathematically developed the normal curve as a continuous approximation to discrete binomial probabilities for large trial sizes. His work remained largely unnoticed until mathematician Karl Pearson rediscovered his original paper in 1924.

  • 1783 (Pierre-Simon Laplace): Laplace applied the normal curve to model the probabilistic distribution of experimental and observational measurement errors.

  • 1809 (Carl Friedrich Gauss): Gauss implemented the curve to analyze astronomical observation errors. In honor of his contributions, the normal distribution is formally designated as the Gaussian distribution.

  • Everyday Terminology: Outside formal mathematical literature, the distribution is commonly referred to as the bell-shaped curve or simply the bell curve.

The Standard Normal Distribution and Z-Score Normalization

• Definition of the Standard Normal Distribution: A unique member of the normal family where the mean is zero and the standard deviation is one:     $\mu = 0$     $\sigma = 1$     $Z \sim N(0, 1)$

• Standardizing Formula ($z$-score Transformation): Any non-standard normal random variable $X \sim N(\mu, \sigma)$ is converted into a standard normal variable $Z$ using the transformation:     $z = \frac{x - \mu}{\sigma}$

  • $x$: The specific raw score or value being evaluated.

  • $\mu$: The population mean of the distribution.

  • $\sigma$: The population standard deviation of the distribution.

• Physical Meaning of a $z$-score:

  • A $z$-score represents the signed distance an individual score $x$ lies from the population mean, measured in units of standard deviation.

  • Positive $z$-score ($z > 0$): The value $x$ lies to the right of (above) the mean.

  • Negative $z$-score ($z < 0$): The value $x$ lies to the left of (below) the mean.

  • Zero $z$-score ($z = 0$): The value $x$ is equal to the mean.

The Empirical Rule (The 68–95–99.7 Rule)

• Prerequisites: The Empirical Rule applies exclusively to distributions that are bell-shaped and symmetric.

• Core Intervals: For any normal random variable $X \sim N(\mu, \sigma)$:

  • $\pm 1\sigma$ Coverage: Approximately $68\%$ of all data values lie within one standard deviation of the mean:         $[\mu - 1\sigma, \mu + 1\sigma] \implies -1 \le z \le +1$

  • $\pm 2\sigma$ Coverage: Approximately $95\%$ of all data values lie within two standard deviations of the mean:         $[\mu - 2\sigma, \mu + 2\sigma] \implies -2 \le z \le +2$

  • $\pm 3\sigma$ Coverage: Approximately $99.7\%$ of all data values lie within three standard deviations of the mean:         $[\mu - 3\sigma, \mu + 3\sigma] \implies -3 \le z \le +3$

  • Almost all observations ($99.7\%$) in a normal distribution fall within three standard deviations of the center.

• Worked Numerical Application ($\mu = 50$, $\sigma = 6$):

  • One Standard Deviation Range ($68\%$):         $-1\sigma = (-1)(6) = -6$         $+1\sigma = (1)(6) = +6$         $\text{Lower limit} = 50 - 6 = 44$         $\text{Upper limit} = 50 + 6 = 56$         Approximately $68\%$ of the data values lie between $44$ and $56$ (corresponding to $z = -1$ and $z = +1$).

  • Two Standard Deviations Range ($95\%$):         $-2\sigma = (-2)(6) = -12$         $+2\sigma = (2)(6) = +12$         $\text{Lower limit} = 50 - 12 = 38$         $\text{Upper limit} = 50 + 12 = 62$         Approximately $95\%$ of the data values lie between $38$ and $62$ (corresponding to $z = -2$ and $z = +2$).

  • Three Standard Deviations Range ($99.7\%$):         $-3\sigma = (-3)(6) = -18$         $+3\sigma = (3)(6) = 18$         $\text{Lower limit} = 50 - 18 = 32$         $\text{Upper limit} = 50 + 18 = 68$         Approximately $99.7\%$ of the data values lie between $32$ and $68$ (corresponding to $z = -3$ and $z = +3$).

Probability Calculations and Cumulative Area Rules

• Probability as Area Under the Curve: In continuous distributions, probability corresponds directly to the area under the density curve over a specified interval.

• Single-Point Probability Zero Property: For any continuous random variable, the probability of obtaining an exact single real number is zero:     $P(X = x) = 0$

  • Consequently, the inclusion or exclusion of boundary endpoints does not affect cumulative probability:         $P(X < x) = P(X \le x)$         $P(X > x) = P(X \ge x)$

• Left-Tail Cumulative Area: The cumulative area to the left of a vertical dividing line at $x$ represents the probability $P(X < x)$.

• Right-Tail Area (Complement Rule): The area to the right of $x$ is computed via the complement rule:     $P(X > x) = 1 - P(X < x)$

• Interval Probability Between Two Points: The probability that a variable falls between values $a$ and $b$ (where $a < b$) equals the cumulative probability up to $b$ minus the cumulative probability up to $a$:     $P(a < X < b) = P(X < b) - P(X < a)$

Standard Normal Worked Case Studies: Thermometer Calibration

• Experimental Context: The Precision Scientific Instrument Company manufactures laboratory thermometers engineered to read $0^\circ\text{C}$ at the freezing point of water. Quality control tests reveal that due to random variation, some thermometers read below $0^\circ\text{C}$ (negative values) and some read above $0^\circ\text{C}$ (positive values).

• Model Assumptions:

  • Mean reading: $\mu = 0^\circ\text{C}$

  • Standard deviation: $\sigma = 1.00^\circ\text{C}$

  • Distribution: Standard normal, $Z \sim N(0, 1)$

  • Sample size: $n = 1$

• Problem 1: Reading Less Than $1.27^\circ\text{C}$:

  • Goal: Determine $P(z < 1.27)$.

  • Standard Normal Cumulative Probability: $P(z < 1.27) = 0.89796 \approx 0.8980$.

  • Complement Probability (Reading Greater Than $1.27^\circ\text{C}$): $P(z > 1.27) = 1 - 0.89796 = 0.10204$.

  • Interval Calculation for $a = 1.27$ and $b = 1.47$: $P(1.27 < z < 1.47) = 0.03126$.

  • Practical Interpretation: The probability of randomly selecting a thermometer with a reading below $1.27^\circ\text{C}$ is $0.8980$; approximately $89.80\%$ of all manufactured thermometers exhibit readings below $1.27^\circ\text{C}$.

• Problem 2: Reading Greater Than $-1.23^\circ\text{C}$:

  • Goal: Determine $P(z > -1.23)$.

  • Left-tail Probability: $P(z < -1.23) = 0.10935$.

  • Right-tail Cumulative Probability: $P(z > -1.23) = 1 - 0.10935 = 0.89065 \approx 0.8907$.

  • Interval Calculation for $a = -1.23$ and $b = 1.50$: $P(-1.23 < z < 1.50) = 0.82384$.

  • Practical Interpretation: The probability of selecting a thermometer reading above $-1.23^\circ\text{C}$ is $0.8907$ ($89.07\%$).

• Problem 3: Reading Between $-2.00^\circ\text{C}$ and $1.50^\circ\text{C}$:

  • Goal: Determine $P(-2.00 < z < 1.50)$.

  • Left-tail Boundary at $a = -2.00$: $P(z < -2.00) = 0.02275 \approx 0.0228$.

  • Right-tail Boundary at $a = -2.00$: $P(z > -2.00) = 0.97725$.

  • Computed Interval Area: $P(-2.00 < z < 1.50) = 0.91044 \approx 0.9104$.

  • Practical Interpretation: If many thermometers are tested at the freezing point of water, $91.04\%$ will output a temperature reading between $-2.00^\circ\text{C}$ and $1.50^\circ\text{C}$.

General Normal Distribution Worked Applications

• Case Study: Transformation of $X \sim N(12, 3)$:

  • Given parameters: Population mean $\mu = 12$, standard deviation $\sigma = 3$, sample size $n = 1$, standard error $\frac{\sigma}{\sqrt{n}} = 3.00000$.

  • Value under analysis: $x = 1$.

  • $z$-score computation:         $z = \frac{x - \mu}{\sigma} = \frac{1 - 12}{3} = \frac{-11}{3} \approx -3.66667 \approx -3.67$

  • Distribution Probabilities:

    • Left-tail probability: $P(x < 1) = 0.00012$

    • Right-tail probability: $P(x > 1) = 0.99988$

    • Interval probability between $a = 1$ and $b = 2.75$: $P(1 < x < 2.75) = 0.00090$

• Case Study: Monthly Weight Loss Program ($X \sim N(5, 2)$):

  • Given parameters: Average monthly weight loss $\mu = 5\,\text{lb}$, standard deviation $\sigma = 2\,\text{lb}$, variable $X = \text{weight lost in pounds per month}$.

  • Scenario A (Loss of $10\,\text{lb}$):

    • Value: $x = 10$

    • $z$-score computation:             $z = \frac{10 - 5}{2} = \frac{5}{2} = 2.5$

    • Interpretation: An individual losing $10\,\text{lb}$ has a weight loss that is $2.5$ standard deviations to the right (above) of the population mean of $5\,\text{lb}$.

    • Associated probabilities: Cumulative left-tail $P(x < 10) = 0.99379$, right-tail exceedance $P(x > 10) = 0.00621$.

  • Scenario B (Gain of $3\,\text{lb}$):

    • A weight gain corresponds to negative weight loss: $x = -3$.

    • $z$-score computation:             $z = \frac{-3 - 5}{2} = \frac{-8}{2} = -4.0$

    • Interpretation: A person gaining $3\,\text{lb}$ is positioned exactly $4$ standard deviations to the left (below) of the population mean of $5\,\text{lb}$.

    • Associated probabilities: Cumulative left-tail $P(x < -3) = 0.00003$, right-tail exceedance $P(x > -3) = 0.99997$.

• Case Study: Statistics Final Exam Examination ($X \sim N(63, 5)$):

  • Given parameters: Mean exam score $\mu = 63$, standard deviation $\sigma = 5$.

  • Task A: Probability of Scoring Higher Than $65$:

    • $z$-score for $x = 65$:             $z = \frac{65 - 63}{5} = \frac{2}{5} = 0.40$

    • Left-tail cumulative probability: $P(X < 65) = 0.65542$

    • Right-tail exceedance probability:             $P(X > 65) = 1 - 0.65542 = 0.34458 \approx 0.3446$

    • Interpretation: The probability that a randomly chosen student scores greater than $65$ is approximately $34.46\%$.

  • Task B: Percentile Rank for a Score of $65$:

    • The percentile rank represents the percentage of scores strictly below the given value.

    • Because $P(X < 65) = 0.65542$, a score of $65$ sits approximately at the $66\text{th}$ percentile.

  • Task C: Probability of Scoring Less Than $85$:

    • $z$-score for $x = 85$:             $z = \frac{85 - 63}{5} = \frac{22}{5} = 4.40$

    • Cumulative probability: $P(X < 85) = 0.99999 \approx 0.9999$

    • Complement probability: $P(X > 85) = 0.00001$

    • Interpretation: Virtually all students ($99.99\%$) score below $85$.

• Case Study: Daily Household PC Entertainment Usage ($X \sim N(2, 0.5)$):

  • Given parameters: Mean usage $\mu = 2\,\text{hours}$, standard deviation $\sigma = 0.5\,\text{hours}$, model $X \sim N(2, 0.5)$.

  • Task: Probability Usage Is Between $1.8$ and $2.75\,\text{hours}$:

    • Lower limit $z$-score ($x_1 = 1.8$):             $z_1 = \frac{1.8 - 2}{0.5} = \frac{-0.2}{0.5} = -0.40$

    • Upper limit $z$-score ($x_2 = 2.75$):             $z_2 = \frac{2.75 - 2}{0.5} = \frac{0.75}{0.5} = 1.50$

    • Cumulative lower probability: $P(X < 1.8) = 0.34458$

    • Interval probability:             $P(1.8 < X < 2.75) = 0.58861 \approx 0.5886$

    • Interpretation: There is a $58.86\%$ probability that a randomly selected household uses the computer for entertainment between $1.8$ and $2.75\,\text{hours}$ per day.

Inverse Normal Computations and Percentiles

• Concept of Inverse Normal Calculations: While standard normal problems evaluate a known score $x$ to determine the corresponding area (probability), inverse normal problems begin with a designated area (probability or percentile) to locate the unknown raw score $k$ or $x_0$.

• Exam Score $90\text{th}$ Percentile Calculation ($X \sim N(63, 5)$):

  • Objective: Identify score $k$ such that $90\%$ of scores are below $k$ and $10\%$ are above $k$:         $P(X < k) = 0.90$

  • Inverse normal output:         $k = 69.40776 \approx 69.4$

  • Interpretation: The $90\text{th}$ percentile exam score is $69.4$. Ninety percent of test takers score below $69.4$, and ten percent score higher than $69.4$.

  • Additional inverse metrics for $P = 0.90$:

    • Upper-tail threshold where $P(X > x_0) = 0.90$: $x_0 = 56.59224$

    • Symmetric central limits: $[54.77573, 63.00000]$

    • Spread parameter: $\mu \pm x_0 \implies 8.22427$

• PC Entertainment Bottom Quartile Calculation ($X \sim N(2, 0.5)$):

  • Objective: Find the maximum number of daily hours representing the bottom quartile ($25\text{th}$ percentile, $k$):         $P(X < k) = 0.25$

  • Inverse normal computation:         $k = 1.66276 \approx 1.66\,\text{hours}$

  • Upper cutoff where $P(X > x_0) = 0.25$: $x_0 = 2.33724\,\text{hours}$

  • Interpretation: The maximum daily entertainment computer usage for the bottom $25\%$ of households is $1.66\,\text{hours}$ per day.

• Smartphone User Age $80\text{th}$ Percentile Calculation ($X \sim N(36.9, 13.9)$):

  • Demographic context: Smartphone owners in the United States aged $13$ to $55+$.

  • Parameters: Mean age $\mu = 36.9\,\text{years}$, standard deviation $\sigma = 13.9\,\text{years}$.

  • Objective: Determine the $80\text{th}$ percentile ($x_0$) such that $P(X < x_0) = 0.80$.

  • Inverse normal computation:         $x_0 = 48.59854 \approx 48.6\,\text{years}$

  • Complementary metrics for $P = 0.80$:

    • Age threshold where $P(X > x_0) = 0.80$: $x_0 = 25.20146\,\text{years}$

    • Central interval boundary: $19.08643$

    • Spread parameter: $17.81357$

  • Interpretation: The $80\text{th}$ percentile is $48.6\,\text{years}$. Exactly $80\%$ of smartphone users in the age bracket $13\text{ to }55+$ are $48.6\,\text{years}$ old or younger, and the remaining $20\%$ are older than $48.6\,\text{years}$.