Continuous Random Variables, Uniform and Normal Distributions, and Applied Poisson Modeling

Fundamentals of Continuous Random Variables

• Definition and Mathematical Support:

  • A continuous random variable takes values along an unbroken continuum or interval rather than over a set of distinct, countable points.

  • The domain may span the entire real number line, $(-\infty, \infty)$, or exist over a bounded subset such as $[0, 1]$, $[1, 10]$, or $[0, 100]$.

• Real-World Phenotypical Continuity vs. Discrete Measurement:

  • Natural phenomena occur continuously regardless of rounding or instrument discretization.

  • A scale reading of $120\,\text{lb}$ represents an instrument-rounded approximation of an exact true weight possessing infinitesimal fractions.

  • Lifetimes of biological organisms (humans, animals) and manufactured items (batteries, light bulbs) represent continuous, non-symmetric decay processes.

  • Physical distances, such as the spatial separation between trees in a forest, fall across continuous spatial intervals.

• Probability Density Functions (PDF):

  • Continuous random variables are modeled by smooth geometric curves, triangles, or rectangular boxes (such as the chi-square distribution, which is bounded at zero and varies smoothly across its domain).

  • The height of a probability density function, denoted $f(x)$, does not represent a probability.

  • Probabilities are strictly defined as areas under the curve over specified intervals.

• Fundamental Mathematical Properties of Continuous Densities:

  • The total area bounded under the curve across the entire support equals $1$:     $\int_{-\infty}^{\infty} f(x)\,dx = 1$

  • The probability of observing an outcome in the interval $[a, b]$ corresponds to the integral:     $P(a \le X \le b) = \int_a^b f(x)\,dx$

  • The probability of observing any single, isolated point is always exactly zero:     $P(X = c) = 0$

  • Point probabilities equal zero because an isolated mathematical point possesses zero width ($\text{width} = 0$), yielding an integrated area of zero.

  • Non-zero probabilities require defining an interval of positive width around a point, such as $[a - 0.1, a + 0.1]$.

  • Inclusion or exclusion of boundary limits does not affect probability values:     $P(X < c) = P(X \le c)$     $P(a < X < b) = P(a \le X \le b) = P(a \le X < b) = P(a < X \le b)$

The Continuous Uniform Distribution

• Overview and Geometric Formulation:

  • Assumes every point within a defined interval $[a, b]$ is equally likely, functioning as an idealized continuous random number generator.

  • Characterized by a flat, rectangular probability density function completely parameterized by its lower limit $a$ and upper limit $b$.

• Probability Density Function (PDF):

  • Evaluated as the reciprocal of the total interval width:     $f(x) = \begin{cases} \frac{1}{b - a} & \text{for } a \le x \le b \\ 0 & \text{otherwise} \end{cases}$

  • For an interval defined on $a = 0$ and $b = 5$, the width is $5 - 0 = 5$, yielding a constant density of $f(x) = \frac{1}{5}$.
    
    - Geometric Interpretation and Area Normalization:
    - Graphically forms a flat, rectangular probability density curve with base length $b - a$ and uniform height $\frac{1}{b - a}$. - The total integrated area under the curve across the entire domain equals exactly $1$, fulfilling the fundamental probability density axiom: $\int_{-\infty}^{\infty} f(x)\,dx = \int_a^b \frac{1}{b - a}\,dx = \left[ \frac{x}{b - a} \right]_a^b = \frac{b - a}{b - a} = 1$ - Density Interpretation vs. Point Probabilities: - The constant height $f(x)$ represents probability density rather than discrete point probability. - Because an isolated mathematical point possesses zero width ($\text{width} = 0$), the probability of observing any single specific point outcome is exactly zero ($P(X = c) = 0$). - Probabilities are strictly defined as areas over intervals of positive width $[c, d] \subseteq [a, b]$: $P(c \le X \le d) = \int_c^d \frac{1}{b - a}\,dx = \frac{d - c}{b - a}$ - Inclusion or exclusion of endpoints does not change the calculated interval probability: $P(c < X < d) = P(c \le X \le d) = P(c \le X < d) = P(c < X \le d)$ - Practical Numerical Demonstration: - For an interval bounded by $a = 0$ and $b = 5$, the total width is $5 - 0 = 5$, yielding a constant density of $f(x) = \frac{1}{5} = 0.20$ for all $x \in [0, 5]$. - To determine the probability of observing an outcome in the sub-interval $[1, 3]$: $P(1 \le X \le 3) = \frac{3 - 1}{5 - 0} = \frac{2}{5} = 0.40$ - Any outcome falling outside the support ($x < 0$ or $x > 5$) has a density of $f(x) = 0$, yielding zero probability outside the interval.

• Cumulative Distribution Function (CDF):

  • Derived by integrating the density function from the lower boundary $a$ to an arbitrary threshold $x$:     $F(x) = \int_a^x \frac{1}{b - a}\,dt = \left[ \frac{t}{b - a} \right]_a^x = \frac{x - a}{b - a}$

  • Complete piecewise specification of the CDF:     $F(x) = \begin{cases} 0 & \text{for } x < a \\ \frac{x - a}{b - a} & \text{for } a \le x \le b \\ 1 & \text{for } x > b \end{cases}$

  • The graph of the CDF remains at $0$ for $x < a$, increases linearly from $0$ to $1$ across $[a, b]$, and remains at $1$ for all $x \ge b$.

  • The geometric midpoint corresponds to a cumulative probability of $0.5$, and a point $75\%$ across the interval corresponds to $0.75$. Cumulative Distribution Function (CDF): - Mathematical Derivation and Integral Formulation: - The Cumulative Distribution Function (CDF), denoted $F(x)$, measures the accumulated probability that the continuous random variable $X$ takes on a value less than or equal to a specified threshold $x$: $F(x) = P(X \le x) = \int_{-\infty}^x f(t)\,dt$ - Because the probability density function of a continuous uniform distribution is non-zero exclusively on the interval $[a, b]$, the lower accumulation limit begins strictly at $a$ ($f(t) = 0$ for all $t < a$). - Integrating the constant density function $f(t) = \frac{1}{b - a}$ from $a$ to $x$ (where $a \le x \le b$) yields: $F(x) = \int_a^x \frac{1}{b - a}\,dt = \left[ \frac{t}{b - a} \right]_a^x = \frac{x - a}{b - a}$ - The result shows that the probability accumulates linearly across the support at a constant rate equal to the reciprocal of the interval width. - Complete Piecewise Specification: - To fully describe probabilities across the entire real number line $(-\infty, \infty)$, the CDF is defined piecewise across three intervals: $F(x) = \begin{cases} 0 & \text{for } x < a \\ \frac{x - a}{b - a} & \text{for } a \le x \le b \\ 1 & \text{for } x > b \end{cases}$ - For outcomes strictly below the support ($x < a$), no probability has accumulated, yielding $F(x) = 0$. - For outcomes within the support ($a \le x \le b$), cumulative probability increases continuously from $0$ to $1$ along a linear slope of $\frac{1}{b - a}$. - For outcomes strictly above the support ($x > b$), all available probability mass has been exhausted, yielding a constant cumulative probability of $F(x) = 1$. - Graphical and Geometric Properties: - Unlike the step functions characteristic of discrete distributions, the CDF of a continuous uniform random variable forms an unbroken, continuous ramp curve. - The curve begins as a horizontal line along the abscissa at height $0$ for $x < a$, smoothly transitions into a line with constant positive slope $\frac{1}{b - a}$ from $(a, 0)$ to $(b, 1)$, and remains horizontal at height $1$ for all $x \ge b$. - By the Fundamental Theorem of Calculus, the derivative of the CDF with respect to $x$ recovers the probability density function wherever the derivative is defined: $\frac{d}{dx}(F(x)) = f(x) = \frac{1}{b - a} \quad \text{for } a < x < b$ - Midpoints, Percentiles, and Quantile Functions: - Cumulative probability increases in exact linear proportion to the fractional distance traversed across the interval $[a, b]$. - The geometric midpoint aligns with the median ($50\text{th}$ percentile or $0.50$ cumulative probability): $F\left(\frac{a + b}{2}\right) = \frac{\frac{a + b}{2} - a}{b - a} = \frac{\frac{b - a}{2}}{b - a} = 0.50$ - The first quartile ($25\text{th}$ percentile) corresponds to one-fourth ($25\%$) of the distance across the support: $F(x) = 0.25$. - The third quartile ($75\text{th}$ percentile) corresponds to three-fourths ($75\%$) of the distance across the support: $F(x) = 0.75$. - The quantile function (inverse CDF, $F^{-1}(p)$) calculates the exact threshold value $x_p$ that accumulates a given target probability $p \in [0, 1]$: $x_p = a + p(b - a)$ - Interval Probability Computation via Cumulative Probabilities: - Evaluating the probability that an observation falls within any sub-interval $[c, d] \subseteq [a, b]$ can be executed directly by evaluating the difference between cumulative distribution function values without re-integrating: $P(c \le X \le d) = F(d) - F(c) = \frac{d - a}{b - a} - \frac{c - a}{b - a} = \frac{d - c}{b - a}$ - Because isolated point probabilities are zero ($P(X = c) = 0$), endpoint inclusions or exclusions do not alter the calculated probability: $P(c < X < d) = P(c \le X \le d) = P(c \le X < d) = P(c < X \le d) = F(d) - F(c)$ - Concrete Numerical Demonstrations: - Demonstration 1 (Standard Support on $a = 0$ and $b = 5$): - The complete piecewise CDF is: $F(x) = \begin{cases} 0 & \text{for } x < 0 \\ \frac{x}{5} & \text{for } 0 \le x \le 5 \\ 1 & \text{for } x > 5 \end{cases}$ - Probability of observing a value less than or equal to the distribution midpoint $x = 2.5$: $P(X \le 2.5) = F(2.5) = \frac{2.5}{5} = 0.50$ - Probability of observing an outcome less than or equal to $x = 4$: $P(X \le 4) = F(4) = \frac{4}{5} = 0.80$ - Interval evaluation for $[1, 3.5]$: $P(1 \le X \le 3.5) = F(3.5) - F(1) = \frac{3.5}{5} - \frac{1}{5} = 0.70 - 0.20 = 0.50$ - Demonstration 2 (Shifted Support on $a = 10$ and $b = 20$): - The total width is $b - a = 20 - 10 = 10$. - Cumulative probability at threshold $x = 13$: $F(13) = \frac{13 - 10}{20 - 10} = \frac{3}{10} = 0.30$ - Determining the $90\text{th}$ percentile threshold ($p = 0.90$): $x_{0.90} = 10 + 0.90 \times (20 - 10) = 10 + 9 = 19$

• Expected Value (Mean) Derivation:

  • Calculated via continuous expectation integration:     $E[X] = \int_a^b x \cdot f(x)\,dx = \int_a^b \frac{x}{b - a}\,dx = \frac{1}{b - a} \left[ \frac{x^2}{2} \right]_a^b = \frac{b^2 - a^2}{2(b - a)}$

  • Factoring the difference of squares $b^2 - a^2 = (b - a)(b + a)$:     $E[X] = \frac{(b - a)(b + a)}{2(b - a)} = \frac{a + b}{2}$

  • The expected value aligns with the physical center of the distribution (e.g., if $a = 10$ and $b = 15$, $E[X] = \frac{10 + 15}{2} = 12.5$).

• Variance and Standard Deviation Derivation:

  • Second moment calculation:     $E[X^2] = \int_a^b \frac{x^2}{b - a}\,dx = \frac{1}{b - a} \left[ \frac{x^3}{3} \right]_a^b = \frac{b^3 - a^3}{3(b - a)} = \frac{b^2 + ab + a^2}{3}$

  • Application of the variance formula $\text{Var}(X) = E[X^2] - (E[X])^2$:     $\text{Var}(X) = \frac{b^2 + ab + a^2}{3} - \left(\frac{a + b}{2}\right)^2 = \frac{b^2 + ab + a^2}{3} - \frac{a^2 + 2ab + b^2}{4}$

  • Finding the common denominator of $12$ (from denominators $3$ and $4$):     $\text{Var}(X) = \frac{4(b^2 + ab + a^2) - 3(a^2 + 2ab + b^2)}{12} = \frac{b^2 - 2ab + a^2}{12} = \frac{(b - a)^2}{12}$

  • Standard deviation expression:     $\sigma = \frac{b - a}{\sqrt{12}}$

  • Dispersion is purely a function of interval width: expanding $b - a$ directly increases the variance.

Normal (Gaussian) Distribution Architecture

• Theoretical Significance and Natural Manifestations:

  • Known as the bell curve or Gaussian distribution, named after mathematician Carl Friedrich Gauss who derived it by observing measurement errors in repeated physical trials.

  • Represents the foundation of modern inferential statistics.

  • Fits natural metrics including physical dimensions (heights and weights), biological variations, and aggregate meteorological rainfall data.

  • Does not naturally fit classroom exam score distributions; curving student exam grades to a normal profile often represents an inappropriate application.

• Parameters and Geometric Controls:

  • Controlled entirely by mean $\mu$ (location) and standard deviation $\sigma$ (scale), with variance $\sigma^2$.

  • Shifting $\mu$ translates the bell curve horizontally along the abscissa without altering its shape.

  • Increasing $\sigma$ widens and flattens the curve, while decreasing $\sigma$ compresses the width and elevates the central peak.

  • Inflection points occur precisely one standard deviation away from the mean at $x = \mu - \sigma$ and $x = \mu + \sigma$.

  • The distribution is strictly symmetric about its center, meaning the mean, median, and mode are identical and equal to $\mu$.

• Mathematical Probability Density Function:

  • Formulated as:     $f(x) = \frac{1}{\sigma \sqrt{2\pi}} e^{-\frac{1}{2}\left(\frac{x - \mu}{\sigma}\right)^2} \quad \text{for } -\infty < x < \infty$

  • The leading scalar $\frac{1}{\sigma \sqrt{2\pi}}$ serves as a normalizing constant ensuring total integrated area equals $1$.

  • The quadratic negative exponent drives the symmetrical exponential decay observed in the tails.

• Calculus Integration Constraints and Computational Methods:

  • The Gaussian kernel $e^{-x^2}$ has no elementary antiderivative and cannot be integrated using standard calculus methods.

  • Probabilities cannot be calculated analytically; they must be solved via numerical approximation methods (such as Riemann sum interpolations), precomputed statistical tables, or statistical software.

Standard Normal Distribution and Z-Score Transformation

• Definition of the Standard Normal Reference ($Z$):

  • A normal distribution with fixed parameters $\mu = 0$ and $\sigma = 1$ (written $Z \sim N(0, 1)$).

  • Symmetrically balanced around $0$, such that $P(Z \le 0) = 0.5$ and $P(Z \ge 0) = 0.5$.

  • Values positioned to the left of the center are negative ($z < 0$); values positioned to the right are positive ($z > 0$).

• The Standardization Transformation (Z-Score):

  • Converts any raw normal variable $X \sim N(\mu, \sigma^2)$ into the standard metric $Z$:     $Z = \frac{X - \mu}{\sigma}$

  • A Z-score quantifies the exact signed number of standard deviations an individual observation $X$ deviates from the population mean $\mu$.

  • Allows any arbitrary normal distribution to map onto a single standardized normal CDF table.

• Probability Computations via Z-Scores:

  • Lower-tail cumulative probabilities are read directly from standard tables:     $P(Z \le z)$

  • Upper-tail probabilities are computed using the complement rule:     $P(Z > z) = 1 - P(Z \le z)$

  • Interval probabilities between two points are found by taking the difference between their respective cumulative values:     $P(z_1 \le Z \le z_2) = P(Z \le z_2) - P(Z \le z_1)$

The Empirical Rule versus Chebyshev's Inequality

• The Empirical Rule (68-95-99.7 Rule):

  • Applies strictly to distributions that exhibit an approximately normal, bell-shaped density.

  • The probability of falling within $1$ standard deviation of the mean is approximately $68.27\%$:     $P(\mu - \sigma \le X \le \mu + \sigma) \approx 0.6827$

  • The probability of falling within $2$ standard deviations of the mean is approximately $95\%$:     $P(\mu - 2\sigma \le X \le \mu + 2\sigma) \approx 0.95$

  • The probability of falling within $3$ standard deviations of the mean is approximately $99.7\%$ (encompassing virtually all observations):     $P(\mu - 3\sigma \le X \le \mu + 3\sigma) \approx 0.997$

• Comparison with Chebyshev's Inequality Bounds:

  • Chebyshev's inequality guarantees that for any probability distribution with finite mean and variance, the probability of falling within $k$ standard deviations ($k > 1$) is at least $1 - \frac{1}{k^2}$:     $P(\mu - k\sigma \le X \le \mu + k\sigma) \ge 1 - \frac{1}{k^2}$

  • For $k = 2$: Chebyshev establishes a lower bound of at least $1 - \frac{1}{4} = 75\%$, whereas the Empirical Rule identifies approximately $95\%$.

  • For $k = 3$: Chebyshev establishes a lower bound of at least $1 - \frac{1}{9} \approx 89\%$, whereas the Empirical Rule identifies approximately $99.7\%$.

  • The Empirical Rule provides tighter bounds than Chebyshev's theorem, but it is invalid unless data normality is verified.

Applications, Critical Values, and Percentiles

• Standardized Probability Calculations:

  • Example with $\mu = 5$ and $\sigma = 2$: Finding $P(X > 7)$.

  • Standardizing the point $X = 7$:       $Z = \frac{7 - 5}{2} = 1.0$

  • By the Empirical Rule, $68\%$ of the distribution lies within $[-1, 1]$, leaving $32\%$ in the outer tails.

  • Due to symmetry, the upper tail contains half of the exterior area:       $P(X > 7) = P(Z > 1) \approx \frac{1 - 0.68}{2} = 0.16$

  • Ground Beef Packaging Example ($\mu = 1.0\,\text{lb}$, $\sigma = 0.1\,\text{lb}$):

  • Evaluating the probability of a randomly selected package weighing between $0.8\,\text{lb}$ and $0.85\,\text{lb}$ ($P(0.8 \le X \le 0.85)$).

  • Calculating corresponding Z-score boundaries:       $Z_1 = \frac{0.8 - 1.0}{0.1} = -2.0$       $Z_2 = \frac{0.85 - 1.0}{0.1} = -1.5$

  • The interval is situated between $-2.0$ and $-1.5$ standard deviations below the mean, yielding an observed probability of less than $5\%$.

• Percentiles and Inverse Normal Calculations:

  • A percentile represents the value below which a given percentage of the distribution falls.

  • Calculating the $99\text{th}$ percentile for ground beef package weights:

  • Identify the critical value $z$ on the standard normal curve where $P(Z \le z) = 0.99$.

  • Reference tables identify this critical value as $Z = 2.33$.

  • Inverting the standardization formula ($X = \mu + Z\sigma$):       $X = 1.0 + (2.33)(0.1) = 1.233\,\text{lb}$

  • Exactly $99\%$ of packages weigh less than $1.233\,\text{lb}$.

• Practical Applications of Normal Percentiles:

  • Retail Shoe Store Inventory Optimization:

  • For women's shoe sizes distributed with mean $\mu = 7$ and standard deviation $\sigma = 2$, stocking inventory across sizes $5$ through $11$ spans two standard deviations on either side of the mean, covering approximately $95\%$ of customer demand.

  • Fisheries Wildlife Regulations:

  • Regulators define minimum weight thresholds, requiring fish smaller than a cutoff $x$ to be released.

  • Setting $x$ to a designated lower percentile protects juvenile fish and ensures sustained reproduction in lakes and reservoirs.

• Software Workflow in StatCrunch:

  • Located under menu path: Calculators $\rightarrow$ Normal.

  • Shows standard parameters $\mu = 0$ and $\sigma = 1$ by default with $P(X \le 0) = 0.5$.

  • Direct interval evaluation: entering $\mu = 1.0$ and $\sigma = 0.1$ allows calculation of upper-tail probabilities such as $P(X \ge 1.2)$ or intervals such as $P(0.7 \le X \le 0.9)$.

  • Percentile derivation: Leaving the value input empty while supplying a target probability calculates the exact threshold.

  • Supplying an upper-tail probability of $0.05$ yields the $95\text{th}$ percentile value of $1.164\,\text{lb}$ (recorded as $1.166\,\text{lb}$ on screen).

  • Supplying a lower-tail probability of $0.10$ yields the $10\text{th}$ percentile value of $0.87\,\text{lb}$, showing that $90\%$ of all packages weigh at least $0.87\,\text{lb}$.

Interactive Problem-Solving: Poisson Distribution Practice (Top Hat)

• Poisson Probability Mass Function:

  • Formula for discrete event counts:     $P(X = x) = \frac{e^{-\lambda} \lambda^x}{x!} \quad \text{for } x = 0, 1, 2, \dots$

  • Zero-event probability simplification: For $x = 0$, $\lambda^0 = 1$ and $0! = 1$, yielding:     $P(X = 0) = e^{-\lambda}$

• Interactive Problem 1: Weekly Accident Rate (Complement Rule Application):

  • Prompt: Accidents occur at a weekly rate of $\lambda = 3$. What is the probability of observing at least one accident this week ($P(X \ge 1)$) ?

  • Calculation:     $P(X \ge 1) = 1 - P(X = 0) = 1 - e^{-3}$

  • Distractor Analysis:

  • The value $3e^{-3}$ represents the probability of exactly one accident ($P(X = 1) = \frac{e^{-3} 3^1}{1!} = 3e^{-3}$).

  • The expression $1 - 3e^{-3}$ represents the complement of exactly one accident ($P(X \ne 1)$), which improperly includes both zero accidents and outcomes greater than one.

  • The correct answer is $1 - e^{-3}$.

• Interactive Problem 2: Parameter Rescaling Across Time:

  • Prompt: If the accident rate is $\lambda = 3$ per week, what is the expected number of accidents observed over a one-month window (assuming a standard $4$-week month)?

  • Calculation: The Poisson rate parameter scales linearly with time exposure:     $\lambda_{\text{month}} = 4 \times \lambda_{\text{week}} = 4 \times 3 = 12$

• Interactive Problem 3: Zero Events Across Rescaled Horizon:

  • Prompt: What is the probability of observing zero accidents across a full one-month interval?

  • Calculation: Substituting the rescaled monthly parameter $\lambda_{\text{month}} = 12$ into the zero-count formula:     $P(X = 0) = e^{-\lambda} = e^{-12}$

• Interactive Problem 4: Insurance Claims and Strict Inequality Parsing:

  • Prompt: An insurance office handles an average of $\lambda = 5$ claims per day. What proportion of days have fewer than $3$ claims?

  • Calculation: The phrase "fewer than $3$" requires a strict inequality ($P(X < 3)$), summing outcomes $x = 0$, $x = 1$, and $x = 2$:     $P(X < 3) = P(X = 0) + P(X = 1) + P(X = 2) = e^{-5} + 5e^{-5} + \frac{5^2}{2!}e^{-5}$

  • Distractor Analysis:

  • Summing outcomes from $0$ to $3$ represents "three or fewer" claims ($P(X \le 3)$), erroneously including $x = 3$.

  • Evaluating solely at $x = 3$ yields the probability of exactly three claims ($P(X = 3)$).

  • Summing outcomes $1$ and $2$ improperly omits days with zero claims ($x = 0$).

• Interactive Problem 5: Exact Event Evaluation and Parameter Positioning:

  • Prompt: For a daily rate of $\lambda = 5$, what is the probability of observing exactly $4$ claims?

  • Calculation: Evaluating the Poisson formula directly at $x = 4$ with $\lambda = 5$:     $P(X = 4) = \frac{e^{-\lambda} \lambda^x}{x!} = \frac{e^{-5} 5^4}{4!}$

  • Distractor Analysis:

  • Formulations reversing $x$ and $\lambda$ (such as $\frac{e^{-4} 4^5}{5!}$ or evaluating $e^{-x}$ instead of $e^{-\lambda}$) represent common structural errors.

  • Cumulative summation expressions represent inequalities ($P(X \le 4)$) rather than the requested exact probability.