Introduction to Statistics and Data Collection Flashcards

Course Mechanics, AI Policy, and Academic Integrity

• Attendance is mandatory for all students, and active attention is strictly required during lectures.
• Due to class size (~40 students), matching every face to a name immediately is challenging, though familiarity develops over time.
• AI Policy: Students are strictly forbidden from uploading any course materials, class content, slides, or intellectual property into artificial intelligence tools for reorganization, searching, or any other purpose.
• Intellectual Property Restrictions: Course materials may not be used or shared for commercial purposes. Selling exams or course materials is strictly prohibited.
• Class Roster & Student Registry: Real-time registration changes occur (e.g., late course additions require manual roll call additions).

Nature and Scope of Statistics

• Structure and Course Progression:
  • Exam 1 Material (Part 1): Focuses on descriptive statistics, organizing, and summarizing data. This is considered the easiest section.
  • Exam 2 Material (Part 2): Covers probability and fundamental inferential principles. It is the most challenging and difficult part of the course, presenting entirely novel concepts for students with no prior statistical background.
  • Final Exam Material (Part 3): The final exam is comprehensive, covering Parts 1, 2, and 3. Part 3 expands inferential concepts to broader applications and different test cases, making it visually more straightforward than Part 2.
• Science vs. Art in Statistics:
  • Pure mathematics yields exact, unique, binary correct or incorrect answers for a given problem.
  • Statistics is both a science and an art. In statistics, multiple answers may be mathematically correct, but many correct answers are completely useless or uninformative.
  • The central mandate in statistics is not merely producing a correct answer, but finding the most useful, informative, and best answer. Submitting a non-informative correct answer does not receive full credit.
  • Statistics deals directly with uncertainty and unknown truths, aiming to derive the best approximation or best guess about reality under limited knowledge.
  • Practical statistical problems are presented as word problems (frequently half a page long) containing abundant real-world data, much of which may be irrelevant or useless. Real-world analysis requires identifying and filtering useful variables from noise.
  • Real-World Example: Contradictory public health claims (e.g., "eggs are good for you" vs. "eggs are not good for you") arise not from mathematical errors, but from researchers applying different analytical judgments and approximations to complex datasets.

Primary Divisions of Statistical Methodologies

• Part 1: Data Collection
  • Foundation of all statistical work; covered conceptually in dedicated study modules.
  • Extremely difficult and time-intensive in practice.
  • If data collection is executed incorrectly or biased, all subsequent organizational, summary, and analytical outputs become completely invalid and useless.
  • Key field application: Experimental design and survey methodologies in biology and life sciences rely heavily on rigorous collection protocols.
• Part 2: Organizing and Summarizing Data (Descriptive Statistics)
  • Encompasses visual representation (tables, graphs, charts) and numerical summaries (means, standard deviations).
  • Focuses on presenting data from alternative analytical perspectives to evaluate when specific measures are appropriate.
• Part 3: Drawing Conclusions (Inferential Statistics)
  • Commences following foundational probability concepts.
  • Focuses on making valid inferences and generalizations about an unobserved entire population based on a smaller observed sample.
  • Involves quantifying uncertainty and establishing confidence levels regarding population parameters.

Fundamental Terminology and Conceptual Definitions

• Population:
  • Definition: The total collection of all objects, individuals, or items under study.
  • Examples: All $2020$ Subaru automobiles produced worldwide, or all registered students in a university course.
• Individual (or Subject):
  • Definition: A single distinct entity or object belonging to the population.
  • Note: An individual is not restricted to a person; it can represent a car, medication dose, plant, animal, or physical unit.
• Sample:
  • Definition: A subset or fraction of the population selected for direct observation and measurement.
• Descriptive Statistics:
  • Definition: Methods involving tables, charts, and numerical calculations used to describe, organize, and summarize data features.
• Inferential Statistics:
  • Definition: Methods used to draw conclusions, estimates, or predictions about an entire population based exclusively on sample data.

Economics and Constraints of Data Collection

• Primary Constraint (Financial Cost):
  • The primary barrier preventing full population measurement is financial cost. Money is the dominant determining factor in statistical study design.
  • Collecting data on millions of individuals requires immense funding for labor and logistics. Given infinite financial resources, obstacles regarding time and manpower can be eliminated by hiring large data collection teams.
• Secondary Constraints (Time and Data Obsolescence):
  • Extended data collection periods run the risk of data becoming obsolete (e.g., taking a year to measure current model automobiles leads to new model year releases or seasonal variations impacting variables like fuel consumption).
• Decennial Census Application:
  • A census represents complete data collection across an entire population.
  • The United States Census is conducted only once every $10$ years because of the monumental financial outlay and labor required, temporarily hiring hundreds of thousands of workers and significantly impacting GDP metrics.

Classification and Types of Variables

• Variable Categorization in Automobile Example (2020 Subaru Gas Consumption Study):
  • Investigated attributes: Miles per gallon ($\text{MPG}$), total mileage driven, fuel tank capacity, vehicle model type, driving environment (city vs. highway miles), cargo/extra weight, and frequency of engine repairs/accidents.
• Qualitative (Categorical) Variables:
  • Definition: Variables describing attributes, qualities, categories, or non-numerical descriptions of an individual.
  • Numerical Qualitative Variables:
  • Numbers that act purely as labels or identifiers where standard arithmetic operations (such as calculating an average) are meaningless.
  • Examples: Vehicle Identification Numbers ($\text{VIN}$), social security numbers, phone numbers, license plate numbers, sports jersey numbers, and birth dates.
  • Test for Qualitative Classification: If computing an arithmetic mean ($\mu$) yields a value that cannot be interpreted logically within the problem context, the variable is qualitative.
  • Classification Subtypes:
  • Nominal: Unordered categories (e.g., eye color, car model type).
  • Ordinal: Categorical variables possessing a logical order or hierarchy (e.g., sleep time ranges $8:00\text{--}10:00\,\text{PM}$, class rankings, academic letter grades $A, B, C, D, F$). Note: Computing arithmetic averages on ordinal ranks is mathematically invalid.
• Quantitative Variables:
  • Definition: Numerical values representing measurable counts or quantities for which standard arithmetic operations (addition, subtraction, averaging) are physically and analytically meaningful.
  • Classification Subtypes:
  • Discrete Quantitative Variables:
    • Take countable values, typically isolated whole integers ($0, 1, 2, 3, \dots$ or $-1, -2, -3, \dots$).
    • Example: Number of engine repairs, number of accidents.
  • Continuous Quantitative Variables:
    • Take uncountable values across a continuous numerical spectrum or interval.
    • Between any two values within a continuous domain, infinitely many intermediate values exist. Observed values depend on measurement precision and rounding limits.
    • Examples: Time, distance, speed, weight, miles per gallon ($\text{MPG}$), total miles driven, city miles.

Methods for Organizing and Summarizing Qualitative Data

• Frequency and Relative Frequency Tables:
  • Frequency Distribution: A table listing all distinct values/categories of a variable alongside the exact counts (frequencies) observed in the sample.
  • Relative Frequency Distribution: A table listing all distinct values alongside their proportional representation relative to the sample total.
  • Formula:     $\text{Relative Frequency} = \frac{\text{Frequency}}{N}$     Where $N$ is the total sample size.
  • Rule of Completeness: A true distribution must account for all possible non-overlapping outcomes across the sample space; the sum of all relative frequencies must equal $1.00$ ($100\%$).
• Numerical Example (Subaru Model Sample, $N = 100$):
  • Crosstrek: Frequency $= 20$, Relative Frequency $= \frac{20}{100} = 0.20$ ($20\%$)
  • Outback: Frequency $= 25$, Relative Frequency $= \frac{25}{100} = 0.25$ ($25\%$)
  • Legacy: Frequency $= 20$, Relative Frequency $= \frac{20}{100} = 0.20$ ($20\%$)
  • Forester: Frequency $= 30$, Relative Frequency $= \frac{30}{100} = 0.30$ ($30\%$)
  • WRX: Frequency $= 5$, Relative Frequency $= \frac{5}{100} = 0.05$ ($5\%$)
  • Total: Frequency $= 100$, Relative Frequency $= 1.00$ ($100\%$)
• Graphical Displays for Qualitative Data:
  • Bar Graphs:
  • Distinct categories are displayed on the horizontal axis ($x$); frequency or relative frequency is displayed on the vertical axis ($y$).
  • Structural Rules: Bars must be of equal width, proportional in height to frequency/relative frequency, and must NOT touch one another.
  • Pareto Charts:
  • A specialized bar graph where categories are arranged along the horizontal axis in strict descending order of frequency (from highest to lowest count).
  • Side-by-Side Bar Graphs:
  • Used to compare categorical distributions across two distinct groups or samples (e.g., comparing grade distributions across two academic semesters).
  • Mandatory Sample Standardization Rule: When comparing two samples of unequal sizes ($N_1 \neq N_2$), comparisons MUST use relative frequencies (percentages) rather than raw frequencies to avoid size-induced misinterpretation.
  • Pie Charts:
  • A circular graph divided into radial sectors representing each category.
  • Sector Angle Calculation Formula:       $\text{Sector Angle} = \text{Relative Frequency} \times 360^\circ$
  • Example (Outback Category with $25\%$ relative frequency):       $\text{Sector Angle} = 0.25 \times 360^\circ = 90^\circ$

Methods for Organizing and Grouping Continuous Quantitative Data

• Class Interval Grouping:
  • Continuous variables cannot be tabulated individually due to infinitely uncountable potential values.
  • Values must be partitioned into contiguous, non-overlapping numerical ranges called classes or class intervals.
  • Structural Rule: Class intervals must possess equal width across the data domain.
• Trade-off in Selecting Class Count and Class Width:
  • Insufficient Class Count (Excessively Wide Classes): Loss of granular detail and data structure (e.g., using $3$ classes of $100000\,\text{miles}$ width over a $0\text{--}300000\,\text{miles}$ domain obscures localized distributions).
  • Excessive Class Count (Excessively Narrow Classes): Over-fragmentation resulting in sparse intervals with $0$ or $1$ observations, obscuring global patterns (e.g., using $30$ classes of $10000\,\text{miles}$ width).
  • Analytical Strategy: Balance class width to maximize pattern clarity without losing essential details or generating empty noise intervals.
• Boundary Handling and Half-Open Interval Definition:
  • To guarantee every continuous observation maps to exactly one class without double-counting boundary values (e.g., a vehicle with exactly $50000\,\text{miles}$), classes are constructed as half-open intervals.
  • Standard Left-Closed, Right-Open Interval Structure:     $\text{Class 1} = [0, 50000)$$\text{Class 2} = [50000, 100000)$$\text{Class 3} = [100000, 150000)$$\text{Class 4} = [150000, 200000)$$\text{Class 5} = [200000, 250000)$$\text{Class 6} = [250000, 300000)$
  • Mathematical Definition: The interval $[a, b)$ contains all real numbers $x$ satisfying $a \le x < b$. An observation where $x = 50000$ is excluded from Class 1 and uniquely included in Class 2.