Introduction to Statistics and Levels of Measurement

Overview of Descriptive and Inferential Statistics

• Descriptive Statistics:

  • Definition: Statistics used to summarize a group of numbers.

  • Core Function: Summarizes and describes the characteristics of a sample or the responses obtained from participants who attend a study or survey.

  • Sample Definition: The specific group of participants being studied (for example, a study sample containing $29$ participant scores).

  • Data Representation: Condenses large, complex datasets into understandable numerical summaries, tables, or figures.

  • Common Examples:

  • Mean ($\text{average}$): Eyeballing a dataset of $29$ scores ranging from $0$ to $5$ might yield an estimated mean in the ballpark of $3.5$.

  • Standard Deviation: Quantifies the dispersion or variability of numbers around the mean.

  • Scope: Descriptive statistics are strictly based on actual numbers obtained from the observed sample.

• Inferential Statistics:

  • Definition: Statistics used to draw conclusions or inferences about a broader population.

  • Root Meaning: Contains the root word "infer," which signifies drawing logical conclusions.

  • Population Definition: The larger group of people to which a researcher intends their study results to apply.

  • Core Function: Enables researchers to make broader assertions and generalizations that extend beyond the specific sample data collected.

Logic of Inferential Statistics: Demonstration and Hypothesis Testing

• Card Demonstration Setup:

  • Materials used: A standard deck of $52$ playing cards, a piece of paper, and a small screwdriver.

  • Demonstration Procedure:

  • A volunteer checks the deck to confirm it is fair and unmanipulated.

  • The volunteer inserts a screwdriver into the side of the deck to choose a random card.

  • The card selected is the Nine of Spades.

  • The card is shown to the audience while hidden from the demonstrator, returned to the deck, and subsequently relocated using the screwdriver.

• Three Potential Explanations / Inferences for the Outcome:

  • Explanation 1: The Effect is Real (Extrasensory Perception / ESP)

  • Assumption: The demonstrator possesses mind-reading or extrasensory perception abilities.

  • Explanation 2: Chance or Luck

  • Probability Calculation: In a standard $52$-card deck, there is only $1$ Nine of Spades. The probability of selecting and identifying this exact card purely by chance is:       $\text{Probability} = \frac{1}{52} \approx 2\%$

  • Decision Rule: Because a $2\%$ ($0.02$) probability is extremely low, chance or luck is considered highly unlikely and is ruled out as the explanation.

  • Explanation 3: Cheating

  • Setup: The outcome occurred because of deliberate manipulation, such as having multiple Nine of Spades cards in the deck or using sleight of hand.

  • Scientific Assumption: In scientific research, cheating is assumed to be excluded from consideration so that researchers can evaluate legitimate hypotheses.

• Core Inferential Reasoning ("Backwards Logic"):

  • Researchers evaluate the probability that chance or luck alone produced the observed results.

  • If the calculated probability of chance or luck is sufficiently low, the chance explanation is ruled out.

  • Upon ruling out chance (and assuming no cheating), the alternative explanation—that the observed effect is real—is accepted as the most plausible conclusion.

Questions & Discussion on Hypotheses

• Null Hypothesis vs. Alternative Hypothesis:

  • Null Hypothesis:

  • Definition: Represents the baseline assumption that there is no effect, no difference, or that outcomes are due entirely to chance or luck.

  • Application: Evaluates the probability that there is no difference between an experimental group and a control group in the broader population.

  • Research (Alternative) Hypothesis:

  • Definition: Represents the hypothesis that the observed effect is real and that a genuine difference exists in the population beyond mere chance.

Levels of Measurement

• Purpose of Measurement Scales:

  • The level of measurement of a variable determines which statistical tests and inferential procedures are mathematically valid to perform.

  • Matching variable types with correct statistical techniques guides appropriate statistical analysis.

• Nominal Scale:

  • Definition: Values reflect names, labels, or qualitative categories (frequently termed categorical variables).

  • Characteristics: Non-numeric in nature.

  • Assigning Numbers: Numeric codes assigned to categories in a dataset (such as assigning numbers $1$, $2$, $3$, or $4$ to categories) are arbitrary codes and do not convert qualitative variables into quantitative ones.

  • Examples:

  • Eye color categories (e.g., blue, brown, hazel, green).

  • Psychiatric diagnostic categories.

• Rank Order Scale:

  • Definition: Values reflect relative rankings or positions within a dataset ($1^{\text{st}}$, $2^{\text{nd}}$, $3^{\text{rd}}$, etc.).

  • Characteristics: Expresses relative standing (whether a case has more or less of an attribute relative to another), but does not quantify the precise distance between ranks.

  • Example (NFL Power Rankings from ESPN.com):

  • Top ranked teams: $1^{\text{st}}$ Los Angeles Rams, $2^{\text{nd}}$ Seattle Seahawks, $3^{\text{rd}}$ Buffalo Bills.

  • Mid-tier teams: Pittsburgh Steelers, Washington Commanders.

  • Limitations: The underlying difference in quality or talent between adjacent ranks varies across the scale. For instance, the difference in ability between the $1^{\text{st}}$ and $2^{\text{nd}}$ ranked entities is often large, whereas the difference between mediocre entities ranked further down (e.g., adjacent mid-tier teams or the $100^{\text{th}}$ vs. $101^{\text{st}}$ runner in a cross-country race) is typically much smaller.

• Equal Interval Scale:

  • Definition: Scale where equal numerical differences between values correspond to equal differences in the actual underlying attribute being measured.

  • Example (Fahrenheit Temperature Scale):

  • The difference between $31^\circ\text{F}$ and $32^\circ\text{F}$ is $1^\circ\text{F}$.

  • The difference between $51^\circ\text{F}$ and $52^\circ\text{F}$ is $1^\circ\text{F}$.

  • The physical change in thermal energy (heat) between $31^\circ\text{F}$ and $32^\circ\text{F}$ is identical to the physical change in thermal energy between $51^\circ\text{F}$ and $52^\circ\text{F}$.

  • Non-Example: Rank order scales fail the equal interval requirement because a rank difference of $1$ unit does not represent a uniform quantity of the underlying property across all levels of the scale.

• Ratio Scale:

  • Definition: Possesses all characteristics of an equal interval scale plus an absolute, true zero point ($0$).

  • True Zero Point: A value of $0$ on the scale corresponds to a complete absence of the physical quantity or attribute being measured.

  • Key Advantage: Enables direct ratio comparisons (e.g., stating one value is "twice as much" as another).

  • Example 1 (Kelvin Temperature Scale):

  • Absolute zero ($0\,\text{K}$, equivalent to approximately $-273^\circ\text{C}$) represents the complete absence of heat.

  • A temperature of $2\,\text{K}$ contains exactly twice as much thermal energy as $1\,\text{K}$ (a ratio of $2:1$).

  • Example 2 (Elapsed Time in a Race):

  • A time of $0\,\text{s}$ represents no elapsed time.

  • Finishing a race in $1\,\text{min}$ is twice as fast as finishing in $2\,\text{min}$ (a ratio of $2:1$).

Data Terminology and Frequency Distributions

• Basic Data Units:

  • Case / Participant: The individual entity or respondent from whom data is gathered.

  • Score: The specific numerical value observed for a given case.

  • Usage: Case, score, and participant are frequently used interchangeably when describing empirical datasets (such as a sample of $29$ scores).

• Frequency:

  • Definition: The total count of cases or scores that possess a specific value within a dataset.

  • Example: In a sample of $29$ scores ranging from $0.0$ to $5.0$, if the score $5.0$ appears exactly once, its frequency is $1$.

• Frequency Distribution:

  • Definition: The pattern showing the number of observed scores across the full range of potential values (e.g., scores ranging from $0$ to $5$).

  • Organization: Displayed in tables or visual figures to summarize how frequently each possible score occurs within a sample.