Module 2: Describing, Exploring, and Comparing Data & The Role of the Forensic Scientist

Role and Operational Framework of the Forensic Scientist

• Fundamental Qualities and Standards of the Forensic Scientist:

  • Forensic scientists must maintain strict impartiality, objectivity, and fairness.

  • All technical workflows, instruments, and testing methods must be formally accredited.

  • Practitioners must adhere strictly to established standard operating procedures (SOPs).

  • Experimental and analytical procedures must yield fully reproducible results.

  • Forensic specialized knowledge must be applied directly to the rigorous interpretation of physical evidence.

  • The primary purpose is to assist and inform the justice system.

• Core Operational Responsibilities:

  • Providing timely intelligence and investigative information to police agencies.

  • Reducing and quantitatively evaluating the uncertainty surrounding forensic examinations.

  • Ensuring forensic services fulfill three essential delivery criteria: validity, cost-effectiveness, and timeliness.

• Investigative vs. Evaluative Roles:

  • Forensic scientists operate within two distinct functional capacities depending on the timing and location of their involvement (e.g., crime scene response versus specialized laboratory analysis).

  • While scientists frequently operate across both domains, distinguishing between them is critical for preventing bias and cognitive contamination (Jackson et al., 2006).

• Detailed Contrast of Roles:

  • Investigator (Investigative Activity):

    • Orientation: Crime-focused.

    • Methodology: Observes the scene and artifacts directly.

    • Purpose: Generates hypotheses and actionable investigative leads.

    • Function: Suggests explanations and offers directional guidance to detectives.

    • Risk: Bears a documented possibility of misleading the investigation if initial working hypotheses are flawed.

    • Output: Provides speculative or preliminary answers.

    • Timeline: Intervenes primarily at the beginning of the investigative process.

  • Evaluator (Evaluative Activity):

    • Orientation: Suspect-focused (evaluating findings relative to competing propositions).

    • Methodology: Maintains strict institutional impartiality; does not act as an advocate for either party.

    • Engagement: Works equally for the prosecution and the defense.

    • Framework: Operates strictly with propositions formally provided by both sides.

    • Function: Evaluates the conditional likelihood of forensic findings given specific propositions.

    • Quality Metrics: Maintains a documented, demonstrably low error rate.

    • Output: Assists the court in reaching informed verdicts without usurping the role of the judge or jury.

    • Timeline: Intervenes at the conclusion of the investigative workflow during formal legal proceedings.

Logical Inference: Deduction and Induction in Forensic Science

• The Foundation of Inference:

  • Inference is the formal logical operation that enables an individual to proceed from established premises to reasoned conclusions.

  • Reasoning in forensic and biological sciences primarily bifurcates into deductive inference and inductive/abductive inference.

  • Everyday forensic reasoning frequently incorporates a dynamic interplay of both modalities.

• Sherlock Holmes and the Misnomer of Deduction:

  • In A Study in Scarlet (Sir Arthur Conan Doyle, 1887), Sherlock Holmes asserts that the science of inference is practical enough to provide his "bread and cheese" through his article The Book of Life.

  • Although literary references describe Holmes as having an inclination for "deduction," an analysis of his reasoning reveals that his conclusions are inductive.

  • Holmes's Inference Regarding Dr. Watson:

    • Observations: Medical gentleman with a military posture; skin tanned on the face but fair at the wrists (indicating tropics rather than natural complexion); haggard appearance (indicative of systemic sickness); stiff, wounded left arm.

    • Holmes's conclusion: Watson served as an army doctor in Afghanistan.

    • Logical assessment: Because an army doctor could have received wounds and illness in other tropical stations, the conclusion is probable rather than mathematically certain, confirming the process as inductive.

• Science, Certainty, and the Fallacy of Exactness:

  • Lay individuals frequently presume that science is exact and provides absolute certainty via deductive logic.

  • The concept of an "exact science" is often considered an oxymoron in forensic interpretation (Evett, 1996).

  • Most empirical scientific inquiries proceed by inferring causes from observed effects, placing induction at the core of forensic evaluation.

• Deductive Inference:

  • Argues from the general law to the particular observation.

  • Starts with a given hypothesis or established causal model and predicts the specific effects that must follow if the premise holds true.

  • Classical Syllogism Example:

    • Premise 1: All humans are mortal.

    • Premise 2: Socrates is human.

    • Conclusion: Therefore, Socrates is mortal.

  • Coin Toss Example (Deduction):

    • Premise 1: If a coin is fair, the probability of heads is $0.5$ and the probability of tails is $0.5$.

    • Premise 2: Probability theory enables the calculation of particular outcome frequencies.

    • Premise 3: The coin is fair.

    • Premise 4: The coin is tossed $10$ times.

    • Conclusion: The exact mathematical probability of observing $3$ heads and $7$ tails can be computed directly using probability theory.

• Key Characteristics of Deduction (Aitken & Taroni, 2004; Goodman, 1999):

  • All information appearing in the conclusion is already implicitly or explicitly contained in the premises; the conclusion adds no novel knowledge beyond the premise.

  • It is logically impossible for the conclusion to be false if the premises are true.

  • Deduction is objective because predictions are always true if the starting hypothesis is true.

  • Limitation: Deduction cannot be used to expand knowledge beyond what is already encapsulated in the hypothesis.

• Inductive Inference:

  • Argues in the reverse direction: proceeds from particular empirical observations to general laws or underlying causes.

  • Evaluates which underlying hypothesis is most tenable based on observed evidence.

  • Sunrise Example:

    • Observation: The sun has risen in the east every morning recorded up to the present.

    • Conclusion: The sun will rise in the east tomorrow.

    • Epistemological status: The conclusion represents a probabilistic inference subject to residual uncertainty; it cannot be known with absolute deductive certainty.

  • Coin Toss Example (Induction):

    • Observation: A coin is tossed $10$ times, resulting in $3$ heads and $7$ tails.

    • Inferential Question: Is the coin fair?

    • Analytical status: Evaluating whether the underlying coin is fair from the empirical outcome requires inductive reasoning.

• Key Characteristics of Induction (Goodman, 1999):

  • The concept of forensic evidence is fundamentally inductive; it represents a metric that reflects back from empirical observations to an unobserved underlying truth.

  • The conclusion contains novel information that extends beyond the premises alone.

  • The truth of the premises does not guarantee the truth of the conclusion; conclusions are probable, never absolute.

  • Advantage: Conclusions reach beyond the raw observations, providing the formal engine for learning new things and building scientific knowledge.

  • Management of uncertainty: Because inductive conclusions are uncertain and probabilistic, probability theory serves as the necessary mathematical tool for managing uncertainty.

• Integrated Medical / Forensic Diagnostic Example:

  • Inductive Workflow (Clinical Diagnosis / Forensic Reconstruction): A physician examines a patient exhibiting symptoms $A$, $B$, and $C$ (e.g., cough, fever, rash, sore throat, elevated blood pressure). The physician reasons from these effects backwards to identify which specific causative disease (Illness 1, Illness 2, or Illness 3) produced the clinical profile.

  • Deductive Workflow (Public Health Intervention / Treatment Allocation): A physician visits a school experiencing an epidemic of a confirmed disease (Illness $A$). Not all infected individuals present identical acute manifestations. Using established epidemiological surveys and pharmacological models, the physician reasons deductively from the known presence of the illness to calculate the required volumetric mass of paracetamol needed to treat the collective fever across the cohort.

Measures of Center

• Fundamental Definitions and Notation:

  • Measure of Center: A representative value located at the center or middle of a numeric data set.

  • $\sum$: Denotes the summation of a set of values.

  • $x$: Variable representing individual data values.

  • $n$: Total count of data values within a sample.

  • $N$: Total count of data values within a population.

  • $\bar{x}$: Pronounced "x-bar," representing the sample mean.

  • $\mu$: Pronounced "mu," representing the population mean.

  • $\tilde{x}$: Pronounced "x-tilde" (also denoted $M$ or $\text{Med}$), representing the sample median.

• Arithmetic Mean:

  • Definition: The measure of center computed by summing all data values and dividing the aggregate sum by the total count of values.

  • Sample Mean Formula:         $\bar{x} = \frac{\sum x}{n}$

  • Population Mean Formula:         $\mu = \frac{\sum x}{N}$

  • Properties of the Mean:

    • Sample means drawn from the same population exhibit lower sampling variability than other measures of center.

    • The calculation incorporates every data value in the set.

    • Non-resistant: Highly sensitive to extreme values or outliers; a single extreme score can alter the mean substantially.

  • Definition of Resistance: A statistical metric is resistant if the presence of extreme outliers does not induce substantial changes in its value.

  • Worked Example (Verizon Smartphone Data Speeds):

    • Raw data (in megabits per second, $\text{Mbps}$): $38.5$, $55.6$, $22.4$, $14.1$, $23.1$

    • Sample size: $n = 5$

    • Summation: $\sum x = 38.5 + 55.6 + 22.4 + 14.1 + 23.1 = 153.7$

    • Computation:             $\bar{x} = \frac{\sum x}{n} = \frac{153.7}{5} = 30.74\,\text{Mbps}$

• Median:

  • Definition: The central value when raw data values are sorted in order of ascending or descending magnitude.

  • Properties of the Median:

    • Resistant measure of center: Extreme outliers do not exert disproportionate leverage on its value.

    • Does not directly incorporate the magnitude of every individual data value.

  • Computational Rules for the Median:

    • Step 1: Arrange data in ascending order.

    • Step 2 (Odd Sample Size): The median is the unique value located in the exact center of the sorted array.

    • Step 2 (Even Sample Size): The median is computed as the arithmetic mean of the two central values.

  • Worked Example 1 (Odd Sample Size, $n = 5$):

    • Raw data speeds ($\text{Mbps}$): $38.5$, $55.6$, $22.4$, $14.1$, $23.1$

    • Sorted array: $14.1$, $22.4$, $23.1$, $38.5$, $55.6$

    • The middle position is the 3rd value: $\text{Median} = 23.1\,\text{Mbps}$.

  • Worked Example 2 (Even Sample Size, $n = 6$):

    • Augmented dataset with a sixth value ($24.5\,\text{Mbps}$): $38.5$, $55.6$, $22.4$, $14.1$, $23.1$, $24.5$

    • Sorted array: $14.1$, $22.4$, $23.1$, $24.5$, $38.5$, $55.6$

    • Central positions are the 3rd and 4th values ($23.1$ and $24.5$

    • Computation:             $\text{Median} = \frac{23.1 + 24.5}{2} = \frac{47.6}{2} = 23.80\,\text{Mbps}$

• Mode:

  • Definition: The data value or values that appear with the greatest frequency.

  • Properties of the Mode:

    • Can be computed for quantitative and qualitative (categorical) data.

    • A dataset can possess one mode, multiple modes, or no mode at all.

  • Classifications:

    • Unimodal: A single value exhibits the highest frequency.

    • Bimodal: Exactly two values share the identical maximum frequency.

    • Multimodal: More than two values share the identical maximum frequency.

    • No Mode: Every value appears with identical frequency (no repeated values).

  • Worked Examples:

    • Dataset: $0.2$, $0.3$, $0.3$, $0.3$, $0.6$, $0.6$, $1.2\,\text{Mbps}$. Frequency of $0.3$ is three; frequency of $0.6$ is two. Mode = $0.3\,\text{Mbps}$.

    • Dataset: $0.3$, $0.3$, $0.6$, $4.0$, $4.0\,\text{Mbps}$. Bimodal with modes of $0.3\,\text{Mbps}$ and $4.0\,\text{Mbps}$.

    • Dataset: $0.3$, $1.1$, $2.4$, $4.0$, $5.0\,\text{Mbps}$. No mode present.

• Midrange:

  • Definition: The value positioned halfway between the maximum and minimum values in the dataset.

  • Formula:         $\text{Midrange} = \frac{\text{maximum data value} + \text{minimum data value}}{2}$

  • Properties of the Midrange:

    • Computationally straightforward to derive.

    • Reinforces that the central tendency can be defined through multiple mathematical conventions.

    • Highly non-resistant; depends solely on the most extreme values, making it susceptible to distortion.

  • Worked Example:

    • Data speeds ($\text{Mbps}$): $38.5$, $55.6$, $22.4$, $14.1$, $23.1$

    • Maximum value = $55.6\,\text{Mbps}$; Minimum value = $14.1\,\text{Mbps}$

    • Computation:             $\text{Midrange} = \frac{55.6 + 14.1}{2} = \frac{69.7}{2} = 34.85\,\text{Mbps}$

• Rounding Conventions for Measures of Center:

  • Mean, Median, Midrange: Round results to one additional decimal place beyond what is present in the raw data values.

  • Mode: Maintain values exactly as observed without rounding, as they originate directly from observed observations.

• Critical Thinking and Inappropriate Applications of Center:

  • Measures of center are mathematically invalid when applied to numbers that do not quantify a continuous or discrete scale.

  • Nominal Identifiers (Zip Codes): Gateway Arch ($63102$), White House ($20500$), Pentagon Air Force Division ($20330$), Empire State Building ($10118$), Statue of Liberty ($10004$). Calculating the mean or median of zip codes yields a meaningless number, as zip codes are arbitrary geographic labels.

  • Ordinal Rankings (University Rankings): Harvard ($2$), Yale ($3$), Duke ($7$), Dartmouth ($10$), Brown ($14$). Ranks express relative sequence rather than uniform intervals or counted amounts; hence, computing a mean is mathematically flawed.

• Mean Derived from a Frequency Distribution:

  • When working with grouped frequency data, individual data points are unknown. Computations assume every value in a given class matches its class midpoint ($x$).

  • Formula:         $\bar{x} = \frac{\sum (f \cdot x)}{\sum f}$         where $f$ is class frequency, $x$ is class midpoint, and $\sum f = n$ .

  • Worked Example (Drive-Through Service Times in Seconds):

*   *Calculation:*

        $\bar{x} = \frac{\sum (f \cdot x)}{\sum f} = \frac{8025.0}{50} = 160.5\,\text{seconds}$ * Caveat: This value is an approximation because class midpoints substitute for actual raw observations.

• Weighted Mean:

  • Definition: The mean calculated when different data values ($x$) contribute with varying degrees of importance or weight ($w$).

  • Formula:         $\bar{x} = \frac{\sum (w \cdot x)}{\sum w}$

  • Worked Example (First-Semester Grade Point Average):

    • Grading quality points: $A = 4$, $B = 3$, $C = 2$, $D = 1$, $F = 0$

    • Course 1: Grade $A$ ($x = 4$), Credits ($w = 3$

    • Course 2: Grade $A$ ($x = 4$), Credits ($w = 4$

    • Course 3: Grade $B$ ($x = 3$), Credits ($w = 3$

    • Course 4: Grade $C$ ($x = 2$), Credits ($w = 3$

    • Course 5: Grade $F$ ($x = 0$), Credits ($w = 1$

    • Sum of weights: $\sum w = 3 + 4 + 3 + 3 + 1 = 14$

    • Sum of products:             $\sum (w \cdot x) = (3 \times 4) + (4 \times 4) + (3 \times 3) + (3 \times 2) + (1 \times 0) = 12 + 16 + 9 + 6 + 0 = 43$

    • Computation:             $\bar{x} = \frac{43}{14} \approx 3.07$

    • Rounding note: While the general rounding rule dictates one extra decimal place ($3.1$), collegiate GPAs are conventionally expressed to two decimal places ($3.07$).

Measures of Variation

• Rounding Rule for Measures of Variation:

  • Round all derived metrics of variation to one decimal place beyond the precision of the raw data.

• Range:

  • Definition: The absolute difference between the highest and lowest values in a dataset.

  • Formula:         $\text{Range} = (\text{maximum data value}) - (\text{minimum data value})$

  • Properties: Highly non-resistant; ignores the internal distribution of the intermediate values.

  • Worked Example:

    • Verizon speeds: $38.5$, $55.6$, $22.4$, $14.1$, $23.1\,\text{Mbps}$

    • $\text{Range} = 55.6 - 14.1 = 41.50\,\text{Mbps}$

• Standard Deviation:

  • Definition: A measure of how much data values disperse or deviate away from the arithmetic mean.

  • Notation:

    • $s$: Sample standard deviation.

    • $\sigma$: Population standard deviation.

  • Properties:

    • $s$ is strictly non-negative ($s \ge 0$). It equals zero if and only if all values are identical.

    • Higher values of $s$ reflect greater dispersion.

    • Non-resistant: Outliers can cause extreme inflations of $s$.

    • Units match the units of the original measurements.

    • The sample standard deviation $s$ is a biased estimator of the population standard deviation $\sigma$.

  • Standard Conceptual Formula (Sample):         $s = \sqrt{\frac{\sum (x - \bar{x})^2}{n - 1}}$

  • Shortcut / Computational Formula (Sample):         $s = \sqrt{\frac{n \left(\sum x^2\right) - \left(\sum x\right)^2}{n (n - 1)}}$

  • Population Standard Deviation Formula:         $\sigma = \sqrt{\frac{\sum (x - \mu)^2}{N}}$

  • Worked Example 1 (Standard Formula, Verizon Speeds):

    • Data ($x$): $38.5$, $55.6$, $22.4$, $14.1$, $23.1$

    • Sample size: $n = 5$, degrees of freedom: $n - 1 = 4$

    • Mean: $\bar{x} = 30.74\,\text{Mbps}$ (used as $30.7$

    • Sum of squared deviations: $\sum (x - \bar{x})^2 = 1083.0520$

    • Computation:             $s = \sqrt{\frac{1083.0520}{4}} = \sqrt{270.7630} \approx 16.45\,\text{Mbps}$

  • Worked Example 2 (Shortcut Formula, Same Data):

    • $n = 5$

    • $\sum x = 153.7$

    • $\left(\sum x\right)^2 = (153.7)^2 = 23623.69$

    • $\sum x^2 = (38.5)^2 + (55.6)^2 + (22.4)^2 + (14.1)^2 + (23.1)^2 = 1482.25 + 3091.36 + 501.76 + 198.81 + 533.61 = 5807.79$

    • Numerator: $5(5807.79) - 23623.69 = 29038.95 - 23623.69 = 5415.26$

    • Denominator: $5(5 - 1) = 20$

    • Computation:             $s = \sqrt{\frac{5415.26}{20}} = \sqrt{270.763} \approx 16.45\,\text{Mbps}$

• Variance:

  • Definition: The measure of variation equal to the square of the standard deviation.

  • Notation:

    • $s^2$: Sample variance.

    • $\sigma^2$: Population variance.

  • Properties:

    • Units are expressed as the square of the original measurement units.

    • Non-resistant: Highly sensitive to outliers.

    • Never negative; equals zero only if all values are identical.

    • The sample variance $s^2$ is an unbiased estimator of the population variance $\sigma^2$.

  • Mathematical Rationale for Dividing by $(n - 1)$:

    • Degrees of Freedom: In a sample with fixed mean $\bar{x}$, only $n - 1$ values can be assigned freely; the final value is constrained.

    • Bias Correction: Dividing by $n - 1$ forces sample variances to center on the population variance $\sigma^2$. Dividing by $n$ results in systematic underestimation of $\sigma^2$.

• Range Rule of Thumb:

  • Interpreting Significant Values:

    • The vast majority (approximately $95\%$  of values lie within $2$ standard deviations of the mean.

    • Significantly low values: $\mu - 2\sigma$ or lower.

    • Significantly high values: $\mu + 2\sigma$ or higher.

    • Values not significant: Between $(\mu - 2\sigma)$ and $(\mu + 2\sigma)$.

• Estimating Standard Deviation from the Range:

  • Rough approximation formula:         $s \approx \frac{\text{Range}}{4}$

• The Empirical Rule (for Bell-Shaped / Normal Distributions):

  • Applies strictly to distributions that are approximately symmetric and bell-shaped.

  • Thresholds:

    • Approximately $68\%$ of all observations fall within $1$ standard deviation of the mean ($\bar{x} \pm s$), with $34.13\%$ on each side.

    • Approximately $95\%$ of all observations fall within $2$ standard deviations of the mean ($\bar{x} \pm 2s$), adding $13.59\%$ to each wing.

    • Approximately $99.7\%$ of all observations fall within $3$ standard deviations of the mean ($\bar{x} \pm 3s$), adding $2.14\%$ to each tail, leaving $0.13\%$ in the outer tails.

• Worked Example (IQ Score Analysis):

  • Distribution: Bell-shaped with mean $\mu = 100$ and standard deviation $\sigma = 15$.

  • Target interval: Between $70$ and $130$.

  • Distance calculation: $2\sigma = 2(15) = 30$.

  • Lower boundary: $100 - 30 = 70$.

  • Upper boundary: $100 + 30 = 130$.

  • Conclusion: The scores $70$ and $130$ fall at exactly $\mu \pm 2\sigma$. Therefore, approximately $95\%$ of all IQ scores fall between $70$ and $130$.

• Coefficient of Variation (CV):

  • Definition: A normalized, dimensionless metric expressing standard deviation relative to the arithmetic mean, stated as a percentage.

  • Sample Formula:         $\text{CV} = \frac{s}{\bar{x}} \times 100$

  • Population Formula:         $\text{CV} = \frac{\sigma}{\mu} \times 100$

  • Rounding Rule: Round the coefficient of variation to one decimal place (e.g., $25.3\%$ .

Measures of Relative Standing and Boxplots

• Standard Scores ($z$ Scores):

  • Definition: The number of standard deviations that a given value $x$ is positioned above or below the mean.

  • Sample $z$ Score Formula:         $z = \frac{x - \bar{x}}{s}$

  • Population $z$ Score Formula:         $z = \frac{x - \mu}{\sigma}$

  • Properties:

    • Expresses relative standing with no units of measurement.

    • If a raw score is lower than the mean, the corresponding $z$ score is negative.

    • Rounding standard: Round $z$ scores to two decimal places.

  • Identification of Statistical Significance via $z$ Scores:

    • Significantly Low: $z \le -2.00$

    • Values Not Significant: $-2.00 < z < 2.00$

    • Significantly High: $z \ge 2.00$

• Worked Example 1 (Comparing Different Scales: Birth Weight vs. Body Temperature):

  • Birth Weight Context: Infant weight $x = 4000\,\text{g}$; sample of $400$ weights has mean $\bar{x} = 3152.0\,\text{g}$ and $s = 693.4\,\text{g}$.         $z = \frac{4000 - 3152.0}{693.4} = \frac{848.0}{693.4} \approx 1.22$

  • Adult Body Temperature Context: Temperature $x = 99^\circ\text{F}$; sample of $106$ adults has mean $\bar{x} = 98.20^\circ\text{F}$ and $s = 0.62^\circ\text{F}$.         $z = \frac{99 - 98.20}{0.62} = \frac{0.80}{0.62} \approx 1.29$

  • Comparison: A body temperature of $99^\circ\text{F}$ is $1.29$ standard deviations above the mean, whereas a birth weight of $4000\,\text{g}$ is $1.22$ standard deviations above the mean. Because the temperature lies farther from its mean, it represents the more extreme value.

• Worked Example 2 (Assessing Clinical Significance of Platelet Count):

  • Dataset lowest platelet count: $x = 75$ (measured in $1000\,\text{cells}/\mu\text{L}$ .

  • Population/Sample parameters: Mean $\bar{x} = 239.4$, standard deviation $s = 64.2$.

  • Computation:         $z = \frac{75 - 239.4}{64.2} = \frac{-164.4}{64.2} \approx -2.56$

  • Conclusion: Because $z = -2.56$ is less than $-2.00$, a platelet count of $75$ is significantly low (thrombocytopenia).

• Percentiles:

  • Definition: Measures of relative location denoted $P_1, P_2, \dots, P_{99}$, dividing sorted data into $100$ distinct groups with approximately $1\%$ of observations per group.

  • Finding the Percentile Corresponding to a Specific Value ($x$):         $\text{Percentile of value } x = \frac{\text{number of values less than } x}{\text{total number of values}} \times 100$         (Round the resulting percentile to the nearest whole integer).

  • Reference Dataset (50 Airport Data Speeds in $\text{Mbps}$, Sorted):

*   *Worked Example (Finding Percentile of $11.8\,\text{Mbps}$):*
    *   Count of values strictly less than $11.8\,\text{Mbps}$: Exactly $20$ values.
    *   Total sample size: $n = 50$
    *   Computation:

            $\text{Percentile} = \frac{20}{50} \times 100 = 40$ * Conclusion: The speed of $11.8\,\text{Mbps}$ corresponds to the 40th percentile ($P_{40} = 11.8\,\text{Mbps}$), meaning it separates the lower $40\%$ of values from the upper $60\%$.

• Converting a Percentile to Its Equivalent Data Value:

  • Variables:

    • $n$: Total count of data values.

    • $k$: Percentile rank in question.

    • $L$: Locator giving position in the sorted list.

    • $P_k$: Value of the $k\text{-th}$ percentile.

  • Locator Equation:         $L = \left(\frac{k}{100}\right) n$

*   *Procedural Decision Rules:*
    *   Case 1: If $L$ is a whole number, the percentile $P_k$ is positioned midway between the $L\text{-th}$ and the $(L + 1)\text{-th}$ values. Calculate $P_k$ by summing the $L\text{-th}$ value and the next value and dividing by $2$.
    *   Case 2: If $L$ is not a whole number, round $L$ upward to the next larger integer. The value of $P_k$ is the element residing at that rounded-up $L\text{-th}$ index.
*   *Worked Example (Finding the 40th Percentile, $P_{40}$, for $n = 50$):*
    *   Calculate locator:

            $L = \left(\frac{40}{100}\right) 50 = 20$ * Evaluation: $L = 20$ is a whole number (Case 1). * Extract values: The 20th value is $11.6\,\text{Mbps}$; the 21st value is $11.8\,\text{Mbps}$. * Interpolate:             $P_{40} = \frac{11.6 + 11.8}{2} = 11.7\,\text{Mbps}$

• Quartiles:

  • Definition: Measures of location denoted $Q_1$, $Q_2$, and $Q_3$, partitioning sorted data into four groups containing approximately $25\%$ of values each.

  • Descriptions:

    • $Q_1$ (First Quartile): Matches $P_{25}$; divides the bottom $25\%$ from the top $75\%$.

    • $Q_2$ (Second Quartile): Matches $P_{50}$ and the median; divides the bottom $50\%$ from the top $50\%$.

    • $Q_3$ (Third Quartile): Matches $P_{75}$; divides the bottom $75\%$ from the top $25\%$.

  • Algorithmic Ambiguity Warning: There is no universal mathematical consensus on a single protocol for deriving quartiles; different software tools and statistical engines frequently generate varying values.

• Auxiliary Range Metrics Defined by Quartiles and Percentiles:

  • Interquartile Range (IQR):         $\text{IQR} = Q_3 - Q_1$

  • Semi-Interquartile Range:         $\text{Semi-IQR} = \frac{Q_3 - Q_1}{2}$

  • Midquartile Range:         $\text{Midquartile} = \frac{Q_3 + Q_1}{2}$

  • $10 - 90$ Quartile Range:         $10-90\,\text{Range} = P_{90} - P_{10}$

• The 5-Number Summary:

  • The five specific summary metrics that describe the distribution:

    1. Minimum value

    2. First Quartile ($Q_1$

    3. Second Quartile ($Q_2$ / Median)

    4. Third Quartile ($Q_3$

    5. Maximum value

  • Worked Example (Using 50 Sorted Airport Speeds):

    • Minimum = $0.8\,\text{Mbps}$

    • First Quartile ($Q_1 = P_{25}$): Located at $L = 0.25 \times 50 = 12.5$, rounded up to 13th value: $Q_1 = 7.9\,\text{Mbps}$

    • Second Quartile ($Q_2$ / Median): Midway between 25th ($13.7$) and 26th ($14.1$) values: $Q_2 = \frac{13.7 + 14.1}{2} = 13.9\,\text{Mbps}$

    • Third Quartile ($Q_3 = P_{75}$): Located at $L = 0.75 \times 50 = 37.5$, rounded up to 38th value: $Q_3 = 21.5\,\text{Mbps}$

    • Maximum = $77.8\,\text{Mbps}$

    • Summary: $0.8$, $7.9$, $13.9$, $21.5$, $77.8\,\text{Mbps}$

• Boxplots and Skewness:

  • Definition: A graphical display of data consisting of a horizontal line extending from the minimum to the maximum value, enclosing a rectangular box spanning from $Q_1$ to $Q_3$, with an internal vertical segment marking the median ($Q_2$ .

  • Construction Procedure:

    1. Determine the 5-number summary.

    2. Plot an axis reflecting measurement units.

    3. Form a box extending from $Q_1$ to $Q_3$.

    4. Insert a vertical dividing line at $Q_2$.

    5. Extend whiskers outward from the box edges to the minimum and maximum data values.

*   *Assessing Skewness:*
    *   A dataset is skewed if it lacks symmetry and its tails extend farther to one side than the other.
    *   A boxplot reflects skewness when the median line is positioned away from the box center or when one whisker is noticeably longer than the other.

• Modified Boxplots and Outlier Detection:

  • Modified boxplots formally flag outliers using the interquartile range criterion.

  • Algorithmic Steps for Outlier Identification:

    1. Determine $Q_1$ and $Q_3$.

    2. Compute the Interquartile Range: $\text{IQR} = Q_3 - Q_1$.

    3. Compute the step threshold: $1.5 \times \text{IQR}$.

    4. Calculate Lower Fence: $Q_1 - (1.5 \times \text{IQR})$.

    5. Calculate Upper Fence: $Q_3 + (1.5 \times \text{IQR})$.

    6. Classify any data point lying below the lower fence or above the upper fence as an outlier.

    7. In modified boxplots, whiskers extend only to the most extreme data values residing within the fences; confirmed outliers are plotted as individual isolated points.