W1 L3 Variable Features, Scales, Variance, and Measurement Quality

Variable Scales & Core Definitions

  • A variable (a.k.a. factor) is any attribute that varies across participants/observations.

    • Defining property: possesses at least two values that can differ from person to person or situation to situation.

    • Four classic scales of measurement determine how those values are represented and interpreted.

1. Nominal Scale

  • Values are names / categories only—labels are arbitrary.

    • Example coding of “Country of Birth”: 1! = !\text{Australia},\; 2! = !\text{Austria},\; 3! = !\text{Azerbaijan} …

    • Numerical difference (e.g.
      3 - 2 = 1) is meaningless; numbers simply stand for categories.

  • Data type: typically categorical / integer codes.

2. Ordinal Scale

  • Has identity + magnitude (order).

    • Values indicate rank; intervals are not guaranteed equal.

    • Example: 7-point Sleep Quality rating.

    • “5/7” signifies better sleep than “3/7”, yet the gap 5-3 = 2 doesn’t necessarily equal the subjective difference between “4” and “2”.

  • Use when order matters but precise distance does not.

3. Interval Scale

  • Possesses identity, magnitude, equal intervals.

    • Differences of 1 unit are the same anywhere along the scale.

    • Can include negative values because zero is arbitrary, not absolute.

    • Classic example: temperature (°C)—17\rightarrow18 is the same change as 24\rightarrow25.

  • Mathematical operations of addition/subtraction are meaningful; ratios are not (because 0 is not true absence).

4. Ratio Scale

  • Includes all four features: identity, magnitude, equal intervals, absolute zero.

    • Zero denotes true absence of the quantity.

    • Example: height (cm).

    • In botanical context, a seedling can truly be 0 cm tall at germination.

    • Allows full range of arithmetic: \frac{200\text{ cm}}{100\text{ cm}} = 2 ⇒ twice as tall.

Four Fundamental Features of Variables

  1. Identity – each value has a unique meaning.

  2. Magnitude – values are ordered from lesser to greater.

  3. Equal Intervals – distance between successive units is constant.

  4. Absolute Zero – a value of 0 indicates none of the attribute.

Feature Matrix (Conceptual)

  • Nominal: Identity only.

  • Ordinal: Identity + Magnitude.

  • Interval: Identity + Magnitude + Equal Intervals.

  • Ratio: Identity + Magnitude + Equal Intervals + Absolute Zero.

Linking Scales to Real-World Examples & Data Types

  • Height – Ratio; continuous real numbers.

  • Country of Birth – Nominal; categorical integer codes.

  • Sleep Quality (1–7) – Ordinal; categorical/integers.

  • Academic Performance (0 = Not Meeting, 1 = Meeting) – Ordinal binary; categorical integers.

Variability & Variance in a Mini Data Set

  • Table with 7 fictitious people recorded on the four variables.

    • Each column header (Height, Country, Sleep Quality, Academic Performance) is a variable.

    • Each row is one participant’s data.

  • Variance captures how individual scores differ from one another within a variable.

    • If every participant had identical height, variance for height =0, but that would contradict the very idea of it being a variable.

  • Terminology:

    • Variability – general concept of differences.

    • Variance – specific statistical quantification of variability.

Measuring Psychological Constructs

  • Psychology often targets latent constructs not directly observable (intelligence, anxiety, optimism, cognition).

    • Cannot be “measured with a ruler.”

  • Necessitates carefully designed instruments, surveys, or tasks.

Quality of a Measure

Reliability (Consistency)

  • Produces similar results under consistent conditions.

  • Low random error → low scatter around true value.

Validity (Truthfulness)

  • Actually measures the intended construct, not something else.

  • “Hits the bull’s-eye.”

Dartboard Illustration (Metaphor)

A. Reliable & Valid – tight grouping around bull’s-eye.
B. Unreliable & Valid – scattered but centered.
C. Reliable & Invalid – tight cluster in wrong quadrant.
D. Unreliable & Invalid – wide scatter, off target.

  • Ultimate goal: high reliability + high validity.

  • “Our data are only as good as the measures we use.”

Course Connections & Future Relevance

  • Concepts of scales, features, variance, reliability & validity underpin later topics in statistics, experimental design, and psychometrics.

  • Understanding data types informs choice of statistical tests (e.g., chi-square for nominal, t-test/ANOVA for interval/ratio).

  • Reliability & validity will reappear in courses on test construction, clinical assessment, and survey design.

Key Takeaways

  • There are four measurement scales, each adding an extra feature over the previous.

  • Identity, magnitude, equal intervals, absolute zero define how numbers map onto reality.

  • Variance is inevitable and essential—without it variables would not vary.

  • Psychological measurement demands attending to reliability and validity; mastering these concepts is critical for sound research.