Normal Distribution: Geometric Properties, Computation Methods, Standardization, and Proof of PDF Legitimacy

Computation Methods and Software Tools for the Normal Distribution

• Historical vs. Modern Approaches to Calculation:

  • Statistical tables were historically required for looking up standard normal distribution probabilities.
  • In modern practice, probability tables are primarily utilized during examinations to prevent cheating.
  • Modern analytical software and computational utilities replace tables for practical and professional use:
  • Minitab: Dedicated statistical analysis package.
  • normCDF (Cumulative Distribution Function): Computes cumulative probabilities over a defined interval from a lower bound to an upper bound given a mean and standard deviation.
  • invNorm / inverse norm CDF: Computes the critical value corresponding to a specified cumulative probability percentile.

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

  • Within One Standard Deviation ($\mu \pm 1\sigma$):
  • Calculated as $P(|X - \mu| < \sigma)$.
  • In normCDF, setting the lower bound to $-1$, the upper bound to $1$, the mean to $0$, and the standard deviation to $1$ yields an area of $0.6826$ ($68.26\%$).
  • Within Two Standard Deviations ($\mu \pm 2\sigma$):
  • Evaluated from Exercise Problem 35 on page 57.
  • In normCDF, setting the lower bound to $-2$, the upper bound to $2$, the mean to $0$, and the standard deviation to $1$ yields an area of $0.9545$ ($95.45\%$).
  • Within Three Standard Deviations ($\mu \pm 3\sigma$):
  • Extending the bounds to three standard deviations ($-3$ to $3$) yields an area of $0.997$ ($99.7\%$).
  • Classroom mnemonics and banter referred to these boundaries colloquially as "STD" ("state treasury and defense") and "STV 65".

Geometric Properties and Parameters of the Normal Distribution

• Distribution Parameters:

  • Location Parameter ($\mu$): The mean determines the central location of the distribution curve along the horizontal axis.
  • Scale Parameter ($\sigma$): The standard deviation governs the width, height, and overall spread of the density curve.

• Curve Geometry and Points of Inflection:

  • The points where the probability density function (PDF) changes concavity (inflection points) occur precisely at one standard deviation from the mean, at $\mu - \sigma$ and $\mu + \sigma$.
  • For the standard normal curve, this change in concavity occurs exactly at $-1$ and $1$.
  • Mathematically, this property can be derived directly by taking the second derivative of the normal probability density function and setting it equal to zero.

• The Standard Normal Distribution ($Z$):

  • Denoted by $Z \sim N(0, 1)$, representing a normal distribution with parameters $\mu = 0$ and $\sigma = 1$.
  • Setting $\mu = 0$ and $\sigma = 1$ collapses the general normal probability density function into the simplified standard normal equation.
  • Scaling utility: If an arbitrary normal distribution has $\sigma = 3$, stepping two standard deviations away requires moving to $-6$ and $+6$. Normalizing to standard deviation steps of length $1$ simplifies comparative analysis.
  • Standard notation uses the Greek letter capital Phi ($\Phi$) or capital $P$ to represent the standard normal cumulative distribution function normCDF.
  • The $Z$-score formula transforms any raw value $X$ into standard deviation units: $Z = \frac{X - \mu}{\sigma}$
  • The standard deviation operates as an intuitive "measuring stick," expressing exactly how many steps an observation lies above or below the mean.

Standard Normal Probability Computations

• Upper and Lower Tail Calculations for Critical Value $Z = 1.25$:
  • Group A (Lower Tail):
  • Calculation: $P(Z \le 1.25)$.
  • Evaluation using normCDF: Setting lower bound to $-20$ or $-25$ (a sufficient numerical proxy for negative infinity rather than entering $-1 \times 10^{99}$) and upper bound to $1.25$ yields: $\Phi(1.25) = 0.8944$
  • Group B (Upper Tail):
  • Calculation: $P(Z > 1.25)$.
  • Applying the complement rule: $P(Z > 1.25) = 1 - \Phi(1.25) = 1 - 0.8944 = 0.1056$
  • Confirming consistency: $0.8944 + 0.1056 = 1.0000$.
  • Group C / Drew (Symmetric Negative Lower Tail):
  • Calculation: $P(Z \le -1.25)$.
  • Because the standard normal curve is strictly symmetric about $0$: $P(Z \le -1.25) = P(Z > 1.25) = 0.1056$
  • Group D (Interval Probability):
  • Calculation over the interval from $-0.328$ to $1.25$: $P(-0.328 \le Z \le 1.25) = \Phi(1.25) - \Phi(-0.328)$
  • Evaluation yields $0.5424$.

Applied Example: Construction Haul Times (Exercise 35c)

• Problem Setup:

  • Context: An analysis of a roping process utilized to protect roads around campus during ongoing construction.
  • Variable: Haul time in minutes ($X$).
  • Distribution parameters: Mean $\mu = 8.46\,\text{minutes}$; standard deviation $\sigma = 0.913\,\text{minutes}$.
  • Objective: Determine the probability that the haul time will be between $8$ and $10$ minutes, denoted $P(8 \le X \le 10)$.

• Standardization via $Z$-Scores:

  • Standardizing the upper limit ($X = 10$): $Z = \frac{10 - 8.46}{0.913} \approx 1.69$
  • Standardizing the lower limit ($X = 8$): $Z = \frac{8 - 8.46}{0.913} \approx -0.50$
  • Interpretation: A haul time of $10$ minutes is $1.69$ standard deviations above the mean, while a haul time of $8$ minutes is $0.50$ standard deviations below the mean.

• Computational Methods Comparison:

  • Method 1 (Standard Normal): Inputting standardized bounds into normCDF(-0.50, 1.69, 0, 1) gives a probability of $0.647$.
  • Method 2 (General Normal): Inputting raw data into normCDF(8, 10, 8.46, 0.913) produces the identical result of $0.647$.
  • Practical rationale: Standardizing to $Z$-scores provides intuitive calibration of distance relative to the mean, enables rapid sanity-checking and estimation, and maintains compatibility with standard distribution tables.

Mathematical Proof of PDF Legitimacy: Evaluating the Gaussian Integral

• Objective:

  • To demonstrate that the standard normal probability density function is mathematically legitimate, the total area under the curve across its entire domain must equal $1$: $\int_{-\infty}^{\infty} \frac{1}{\sqrt{2\pi}} e^{-\frac{x^2}{2}}\,dx = 1$
  • This requires proving that the definite integral of the exponential term evaluates to $\sqrt{2\pi}$: $\int_{-\infty}^{\infty} e^{-\frac{x^2}{2}}\,dx = \sqrt{2\pi}$

• Analytical Problem:

  • The function $e^{-\frac{x^2}{2}}$ possesses no elementary antiderivative.
  • While a Taylor series expansion could be integrated term by term, it does not provide a direct closed-form evaluation over infinite limits.

• Multivariable Calculus Strategy (Polar Coordinates Transformation):

  • Step 1: Squaring the Integral:
  • Let $I = \int_{-\infty}^{\infty} e^{-\frac{x^2}{2}}\,dx$.
  • Multiply the integral by an identical copy of itself using an independent dummy variable $y$: $I^2 = \left(\int_{-\infty}^{\infty} e^{-\frac{x^2}{2}}\,dx\right) \left(\int_{-\infty}^{\infty} e^{-\frac{y^2}{2}}\,dy\right)$
  • Step 2: Combining into a Double Integral:
  • Because $x$ and $y$ are completely independent, the product can be rewritten across the entire Cartesian plane $\mathbb{R}^2$: $I^2 = \int_{-\infty}^{\infty} \int_{-\infty}^{\infty} e^{-\frac{x^2 + y^2}{2}}\,dx\,dy$
  • Step 3: Conversion to Polar Coordinates:
  • Apply polar substitutions: $x = r\cos(\theta)$$y = r\sin(\theta)$$x^2 + y^2 = r^2$
  • The differential area element converts via the Jacobian: $dx\,dy = r\,dr\,d\theta$
  • Integration bounds over the entire plane: radius $r$ spans from $0$ to $\infty$; angle $\theta$ spans from $0$ to $2\pi$: $I^2 = \int_{0}^{2\pi} \int_{0}^{\infty} e^{-\frac{r^2}{2}} r\,dr\,d\theta$
  • Step 4: Integration with Respect to $r$ (via $u$-substitution):
  • Let $u = -\frac{r^2}{2}$, which gives $du = -r\,dr$, or $r\,dr = -du$.
  • Integrating the radial component: $\int_{0}^{\infty} r e^{-\frac{r^2}{2}}\,dr = \left[-e^{-\frac{r^2}{2}}\right]_{0}^{\infty} = \lim_{b \to \infty} \left(-e^{-\frac{b^2}{2}}\right) - (-e^0) = 0 - (-1) = 1$
  • Step 5: Integration with Respect to $\theta$:
  • Evaluating the angular component across $[0, 2\pi]$: $I^2 = \int_{0}^{2\pi} 1\,d\theta = 2\pi$
  • Step 6: Deriving the Final Normalizing Constant:
  • Taking the principal square root of both sides: $I = \int_{-\infty}^{\infty} e^{-\frac{x^2}{2}}\,dx = \sqrt{2\pi}$
  • Dividing by the factor $\sqrt{2\pi}$ confirms that the area under the normal density curve equals exactly $1$, proving the PDF is mathematically legitimate.

Questions and Classroom Discussion

• Audience Query on Concavity:

  • Question: Does concavity only change at the first standard deviation, or does it change elsewhere?
  • Response: The concavity of the normal curve changes strictly at $\mu - \sigma$ and $\mu + \sigma$ (at $-1$ and $1$ standard deviations). Setting the second derivative of the PDF to zero verifies that these are the only inflection points on the curve.

• Audience Query on Standard Normal Formulation:

  • Question: What is the fundamental difference between the general normal formula and the standard normal formula, and where does the standard normal formula originate?
  • Response: The parameters of a general normal distribution are specified after the semicolon as $N(\mu, \sigma^2)$. By setting the mean $\mu = 0$ and the standard deviation $\sigma = 1$, the algebraic terms simplify and collapse into the standard normal density equation, representing distance directly in standard deviation units.

• Audience Query on Utilizing $Z$-Scores vs. Raw Inputs:

  • Question: Why should students perform the extra step of standardizing into a $Z$-score if calculators can compute probabilities directly using raw mean and standard deviation inputs?
  • Response: Standardizing provides immediate intuition regarding how extreme a value is in terms of standard deviation steps away from the mean, enables rapid estimation to verify calculator outputs, and aligns with standardized statistical tables.

• Audience Query on the Double Integral Setup:

  • Question: Why was the single integral multiplied by itself using a different variable, and does the second line equal the first?
  • Response: Multiplying the single integral by itself using dummy variable $y$ creates an equivalent square of the original integral ($I^2$). Because $x$ and $y$ are independent, combining them into a double integral allows the algebraic sum $x^2 + y^2$ in the exponent to be converted into polar coordinates ($r^2$), enabling an analytical solution via $u$-substitution that is impossible in single-variable Cartesian coordinates.

• Course Administration, Banter, and Logistics:

  • Reference to multivariable calculus techniques covered in Calculus III under Dr. Robin.
  • Classroom references to popular culture, including "Donkey Kong" and gaming remarks ("X Maxwell").
  • Homework directive: Students must work on Section 4.3 (with prior review material in Section 4.2).
  • Recorded instructional video available covering this calculation; recommended for viewing at $2\times$ speed due to pacing.
  • Student discussion regarding evening exams held the prior night from 6:00 PM to 9:00 PM (referred to as "6 to 9 and 69"), scheduling conflicts at 3:00 PM or 4:00 PM involving visiting/transferred parents, and morning email inquiries regarding assignment extensions.