Statistical Analysis and Biostatistics: Evaluative Roles, Logical Inference, and Descriptive Statistics
Core Roles and Operational Frameworks of the Forensic Scientist
Fundamental Ethical and Methodological Mandates:
Forensic scientists must remain strictly unbiased, objective, and fair across all inquiries.
Analytical methodologies must adhere to accredited techniques and follow standard operating procedures (SOPs).
Experimental testing and casework analyses must achieve reproducible results.
Forensic practitioners utilize specialized scientific knowledge to interpret physical evidence correctly and aid the administration of justice.
Service delivery requires valid, cost-effective findings delivered in a timely fashion.
Categorization of Forensic Activities:
Forensic science operations are bifurcated into two foundational roles: investigative activities and evaluative activities.
Practitioners frequently alternate between both functions depending on the case stage (e.g., crime scene investigation versus courtroom evaluation), making the distinction between an investigator and an evaluator essential (Jackson, Jones, Booth, Champod, & Evett, 2006).

Investigative Role (The Forensic Investigator):
Focuses primarily on the crime scene, physical clues, and reconstructive questions ("What happened? What is this?").
Observes raw scene data, generates working hypotheses, and establishes tactical leads for police inquiries.
Suggests alternative explanations and provides technical guidance to law enforcement.
Carries an inherent operational risk of generating misleading explanations if hypotheses are accepted prematurely without comprehensive testing.
Provides direct answers and acts primarily at the beginning of the criminal justice process.
Evaluative Role (The Forensic Evaluator):
Focuses on specific suspects, formal allegations, and competing legal claims ("Given the activity, how likely is the evidence?").
Operates with complete impartiality, serving both prosecution and defense.
Operates within predefined propositions provided by the legal context (: prosecution hypothesis; : defense hypothesis).
Evaluates findings conditionally under competing propositions rather than asserting absolute factual guilt or innocence.
Demonstrates low, quantitatively documented error rates.
Assists the court in taking informed decisions and operates primarily at the end of the criminal justice process.
Epistemology and Forms of Scientific Inference
Logical Foundations of Inference:
Inference is the formal logical operation that enables an individual to transition from premises to valid conclusions.
Scientific reasoning relies on structured tools to guarantee that inferences remain rational, reproducible, and robust against cognitive bias.
Human reasoning frequently utilizes a dynamic mixture of deduction and induction (or abduction).
The Myth of Absolute Certainty in Science:
Lay individuals commonly assume that science is an "exact" enterprise that yields absolute certainty through deductive proofs; however, equating empirical science with absolute certainty is an oxymoron (Evett, 1996).
Natural and forensic sciences are inherently empirical and probabilistic, depending primarily on inductive inference (inferring causes from observed effects).
Deductive Inference:
Moves systematically from the general law or overarching cause to the particular observation or effect.
Operates such that if the general premises are true, the conclusion is guaranteed to be true.
All information appearing in the conclusion is already implicit within the premises; deduction does not expand knowledge beyond what is already stated.
Classical Syllogism Example (Aitken & Taroni, 2004):
Premise 1: All humans are mortal.
Premise 2: Socrates is human.
Conclusion: Therefore, Socrates is mortal.
Coin Tossing Deductive Formulation:
Premise 1: If a coin is fair, the probability of heads is and the probability of tails is .
Premise 2: Probability theory enables the calculation of particular sequences and outcomes.
Premise 3: The coin is fair.
Premise 4: The coin is tossed times.
Conclusion: The exact probability of observing heads and tails can be computed deductively from probability theory.
Medical Deduction Scenario:
A physician visits a school experiencing an epidemic illness.
Using established surveys and known epidemiological parameters regarding the proportion of individuals who develop a high temperature, the physician deduces the exact quantity of paracetamol required to treat the expected fever cases.

Inductive Inference:
Moves from particular observations or effects back to underlying causes, hypotheses, or general laws.
The concept of physical evidence is inherently inductive; observations reflect back toward an underlying truth.
Inductive inferences enlarge human knowledge: the conclusion asserts propositions about unobserved states of nature that are broader than the immediate premises.
The truth of the premises does not guarantee the truth of the conclusion; conclusions are probable rather than certain.
The Sherlock Holmes Canonical Example (Sir Arthur Conan Doyle, 1887, A Study in Scarlet):
In the discussion concerning the essay "The Book of Life," Watson claims that inferring a person's life history from observation is impractical rubbish; Holmes retorts that his entire career depends on the science of inference for his "bread and cheese."
Holmes infers Dr. Watson's service in Afghanistan by synthesizing particular observational effects:
Professional bearing: Medical air paired with a military demeanor.
Sun exposure: Face deeply tanned, but wrists pale (indicating the tropics, not European sun).
Health status: Haggard, sunken face (reflecting severe illness and hardship).
Physical trauma: Left arm stiff and held unnaturally (indicating a localized physical injury).
Holmes's reasoning: An English army physician suffering severe hardship and injury in the tropics points to Afghanistan. This is an inductive inference: the premises support the conclusion with high likelihood, but do not provide mathematical certainty.
Astronomical Induction Example:
Observation: The sun has risen in the east every morning recorded in history.
Conclusion: The sun will rise in the east tomorrow.
Epistemic state: Highly probable, yet fundamentally an inductive projection subject to residual uncertainty.
Coin Tossing Induction Formulation:
Observation: A coin is tossed times, yielding heads and tails.
Inductive Conclusion / Question: Is the coin fair? (One induces the physical nature of the coin from observed data).

Medical Induction Scenario:
A physician examines a patient exhibiting symptoms (effects) such as cough, fever, rash, sore throat, and elevated blood pressure.
The physician must work backwards from observed clinical effects to infer the most tenable causal pathology (Illness 1, Illness 2, or Illness 3).
Managing Uncertainty:
Because inductive conclusions in science and forensics are probable, uncertainty is unavoidable.
Probability theory serves as the primary scientific instrument to quantify, manage, and communicate uncertainty in evaluative findings.
Measures of Center
Definitions and Notational Standards:
A measure of center is a representative numerical value locating the central or middle point of a data set.
denotes the summation operator across a collection of data values.
represents an individual variable or observed data value.
denotes the total number of data values within a sample.
denotes the total number of data values comprising a complete population.
The Mean (Arithmetic Mean):
The measure of center found by summing all values and dividing by the total count:
Sample Mean (, pronounced "x-bar"):
Population Mean (, pronounced "mu"):
Properties of the Mean:
Sample means drawn from the identical population exhibit less sample-to-sample variability than other measures of center.
Integrates every individual data point directly into its computation.
Non-resistant: A single extreme value or outlier alters the mean substantially.
Definition of Resistance:
A statistic is classified as resistant if the presence of extreme outliers does not induce substantial numerical changes.
Worked Example (Verizon Smartphone Airport Data Speeds):
Five consecutive data speeds (in Mbps): , , , ,
Summation:
Mean:
The Median:
The measure of center representing the middle value when the raw observations are arranged in order of magnitude (ascending or descending).
Common notations: (pronounced "x-tilde"), , or .
Properties of the Median:
Highly resistant to extreme observations; large outliers do not shift the median.
Does not directly incorporate the magnitude of every individual data value.
Calculation Rules:
If sample size is odd, the median is the value located in the exact middle of the sorted array.
If sample size is even, the median is the arithmetic mean of the two central numbers in the sorted array.
Worked Example with Odd Number of Values ():
Raw data: , , , ,
Sorted array: , , , ,
Middle position ( value):
Worked Example with Even Number of Values ():
Array augmented with a sixth value (): , , , , ,
Central values: and
Median calculation:
The Mode:
The value or values that appear with the greatest numerical frequency.
Properties of the Mode:
Can be applied to both qualitative (categorical/nominal) and quantitative data.
A dataset can possess no mode, a single mode, two modes (bimodal), or more than two modes (multimodal).
Worked Examples:
Sprint download speeds (in Mbps): , , , , , , . The value appears three times; hence, .
Bimodal data speeds: , , , , . The values and both recur twice; the dataset is bimodal with modes and .
No mode: , , , , . Every value is distinct; no mode exists.
The Midrange:
The measure of center located halfway between the minimum and maximum data values:
Properties and Pedagogical Utility:
Extremely sensitive to extreme values; non-resistant.
Simple to compute and illustrates that "center" can be defined in multiple ways.
Delineating the midrange prevents common confusion with the median.
Worked Example:
Data values: , , , ,
Round-Off Rules for Measures of Center:
For the mean, median, and midrange: Retain one more decimal place than is present in the raw data.
For the mode: Report the value without rounding, keeping it identical to the original observation.
Critical Thinking on Non-Meaningful Averages:
Arithmetic centers should not be computed for numbers that serve as categorical labels or ordinal ranks.
Zip Code Example: The zip codes of the Gateway Arch (), White House (), Pentagon Air Force division (), Empire State Building (), and Statue of Liberty () denote geographic coordinates, not counts or magnitudes. Computing a mean zip code is meaningless.
University Ranks Example: U.S. News & World Report ranks for Harvard (), Yale (), Duke (), Dartmouth (), and Brown () reflect relative ordering without uniform intervals. The mean rank does not quantify actual educational differences.
Mean from a Grouped Frequency Distribution:
Approximates the sample mean by assuming all observations within a class interval equal the class midpoint (): Where denotes class frequency and .
Worked Example (Service Times in Seconds):
Time Interval (seconds) | Frequency () | Class Midpoint () | Product () |
|---|---|---|---|
Totals | — |
Mean computation: .
Weighted Mean:
Utilized when individual data values () carry differing assigned weights ():
Worked Example (Grade-Point Average Calculation):
Course grades and assigned credits: Grade A in 3-credit course, Grade A in 4-credit course, Grade B in 3-credit course, Grade C in 3-credit course, Grade F in 1-credit course.
Quality point grading scale: , , , , .
Assigned weights: ; Values: .
Sum of weights:
Weighted product sum:
GPA computation: (standard GPA practice rounds to two decimal places; rounding rule alone would dictate ).
Measures of Variation
General Round-Off Rule for Dispersion:
Carry one more decimal place than is present in the raw data.
Range:
The absolute difference between the extreme ends of the distribution:
Properties: Non-resistant; ignores the internal dispersion of the intermediate data.
Worked Example: Verizon speeds (, , , ,
Sample Standard Deviation ():
Quantifies the average dispersion of individual observations around the arithmetic mean.
Standard Definitional Formula:
Shortcut Computational Formula (used by calculators and statistical software):
Properties of Standard Deviation:
Non-negative (). Equals zero only when every value in the dataset is identical.
Increases markedly with greater variability or when outliers are introduced.
Expressed in the identical units of measurement as the original observations.
The sample standard deviation is a biased estimator of the population standard deviation .
Worked Example via Definitional Formula:
Data: , , , , ; ,
Sum of squared deviations:
Degrees of freedom:
Standard deviation:
Worked Example via Shortcut Formula:
Data sums: , ,
Numerator:
Denominator:
Standard deviation:
Degrees of Freedom and Rationale for Dividing by :
Only values in a sample can be assigned without constraint; once the mean is fixed, the final observation is mathematically predetermined.
Division by makes the sample variance an unbiased estimator that centers directly on the population variance .
Dividing by causes systematic underestimation of the population variance.
Population Standard Deviation ():
Calculated across an entire population using total size :
Variance (Sample Variance and Population Variance ):
The square of the standard deviation.
Measured in units squared (e.g., , ).
Highly sensitive to outliers; strictly non-resistant.
The sample variance is an unbiased estimator of .
Range Rule of Thumb:
Understanding Dispersion: Roughly of sample values lie within standard deviations of the mean ().
Identifying Statistical Significance:
Significantly low values: or lower.
Significantly high values: or higher.
Values not significant: Between and .
Rough Standard Deviation Estimation:
The Empirical Rule (for Bell-Shaped / Normal Distributions):
Applies exclusively to distributions that are approximately normal and bell-shaped:
Approximately of all observations fall within standard deviation of the mean ().
Approximately of all observations fall within standard deviations of the mean ().
Approximately of all observations fall within standard deviations of the mean ().

Exact Sectional Percentages:
Between and : each side (total ).
Between and : each side.
Between and : each side.
Tails beyond : each side.
Worked Example (Standardized IQ Scores):
Parameter set: , .
Determine the percentage of IQ scores falling between and :
Deviation check:
Boundaries: and
Conclusion: The interval represents exactly , capturing approximately of all IQ scores.
Coefficient of Variation (CV):
Measures relative dispersion independent of absolute scale, expressed as a percentage:
Sample:
Population:
Rounding: Standardized to one decimal place (e.g., ).
Measures of Relative Standing and Boxplots
Standardized z Scores:
Measures the number of standard deviations an individual score is positioned above or below the mean:
Sample:
Population:
Properties of z Scores:
Dimensionless numbers without measurement units.
Negative values indicate observations below the mean; positive values indicate observations above the mean.
Rounded conventionally to two decimal places.
Significance Determination via z Scores:
Significantly low:
Significantly high:
Not significant:

Comparative Extremity Example: Birth Weight vs. Body Temperature:
Problem: Determine which value is more extreme relative to its respective cohort:
Newborn baby weight of (sample , ,
Adult body temperature of (sample , ,
Computation of z scores:
Baby weight:
Body temperature:
Interpretation: The adult body temperature is standard deviations above its mean, whereas the birth weight is standard deviations above its mean. Because , the adult body temperature of is slightly more extreme than the infant birth weight.
Clinical Significance Example: Platelet Count Analysis:
Clinical context: Minimum recorded platelet count is (measured in ) within a distribution where and .
Standardization:
Evaluative outcome: Because , a platelet count of is significantly low (clinically termed thrombocytopenia).
Percentiles ():
Percentiles are measures of location denoted , dividing a sorted dataset into subsets containing approximately of the data each.
Finding the Percentile of a Data Value : (The result is rounded to the nearest whole integer).
Worked Example with Sorted Airport Download Speeds ():
Dataset:
Target value:
Number of values strictly less than :
Percentile calculation:
Interpretation: represents the percentile (), separating the bottom of download speeds from the top .
Converting a Percentile to Its Corresponding Data Value:
Operational parameters:
: Total count of values in the dataset.
: Target percentile in question.
: Locator index representing item position in the sorted list.
: Value of the percentile.
Formula for Locator:

Algorithmic Decision Paths:
If is a whole number: The percentile value is located midway between the value and the subsequent value in the sorted dataset:
If is not a whole number: Round upwards to the nearest higher whole number. The percentile value is the value from the lowest.
Worked Example (Finding when ):
Locator calculation:
Because is a whole number, identify the and observations in the sorted array:
value =
value =
Midpoint calculation:
Quartiles ():
Quartiles divide sorted data into four equal groups containing approximately of observations each:
First Quartile (): Identical to ; separates the bottom of sorted data from the top .
Second Quartile (): Identical to and the sample median; divides the bottom from the top .
Third Quartile (): Identical to ; separates the bottom from the top .
Methodological note: No single universal mathematical procedure defines quartiles; distinct software packages use varying interpolation algorithms.
Summary Statistics Derived from Quartiles and Percentiles:
Interquartile Range:
Semi-interquartile Range:
Midquartile Range:
Quartile Range:
The 5-Number Summary:
The structural foundation for exploratory data analysis, comprising:
Minimum value
First quartile ()
Second quartile (, the median)
Third quartile ()
Maximum value
Worked Example from the 50 Airport Data Speeds:
Minimum =
First Quartile (): Evaluated at ( value) =
Median (): Evaluated at (midway between value and value ) =
Third Quartile (): Evaluated at ( value) =
Maximum =
Resulting 5-Number Summary:
Boxplots (Box-and-Whisker Diagrams):
A graphical representation constructed directly from the 5-number summary:
A horizontal axis spanned by a line extending from the minimum to maximum data value.
A central box (rectangle) extending from to .
A vertical line segment drawn through the box at the median ().

Assessment of Skewness via Boxplots:
A distribution is skewed if it lacks symmetry and extends further to one side.
Right (positive) skewness is indicated when the right whisker and the right half of the box ( to ) are substantially longer than their left counterparts.
Formal Outlier Identification in Modified Boxplots:
Step 1: Compute the interquartile range:
Step 2: Compute the outlier fence factor:
Step 3: Establish inner fences:
Lower boundary:
Upper boundary:
Step 4: Classify any observation lying beyond either inner fence as a statistical outlier.