Unit 5

Learning Objectives
- Understand the difference between constructs and operational variables
- Identify the three most common forms of measurement in psychological science
- Identify and describe three scales of measurement
- Identify and describe three types of reliability
- Identify and describe four components of contrast validity

Constructs vs. Operational Variables
- Happiness - not directly observable (however, it exists)
  - Construct for conceptual variable
- To study constructs, we make inferences about them based on variables we can observe, which we call operational variables.
  - A single construct can have nearly limitless operational variables, and which one we choose will limit our measurement options.
Most common forms of measurement
- Self-Report Measurements
  - Self-report is undoubtedly the most common form of measurement in psychological science. Why? Well, it’s cheap, it’s easy, and there are some things about you that only you know.
- Observational Measurements
  - Remember the car experiment where they observed the type of car Porsche car had bad habit drivers and normal cars had good mannered dirvers
    - Ethics of Observation
      - Normally, informed consent. You can observe without consent if the behavior is publicly visibl, and no identifying information is recorded.
- Physiological Measurements (psychophys)
  - Body part measurements such as the blood pressure or the brain imaging
Scales of Measurement
- All variables, by definition, have at least two levels. Not only do different variables have different numbers of levels, but they also have different measurement scalest. If I asked you what kind of car you drive, you would give me a car brand. You would give me a number if I asked you how tall you are. The former uses a categorical scale, whereas the latter uses a quantitative scale. Let’s discuss this.
- Categorical Variables
  - Have no numerical value
  - No fixed relationship
  - Can be coded numerically, but those numbers are arbitrary
- Quantatiative Variable
  - Ordinal Scale (rankings)
    - levels represent a ranked order, and distances between levels are not equal
      - Sports rankings, Olympic medals, US News college rankings, top 10 lists
  - Interval Scale
    - A quantitative measurement scale that has no “true zero,” and in which the numerals represent equal intervals (distances) between levels
      - Variables such as temperature (except for Kelvin) can go below zero, so the “zero” point is arbitrary. Usually, this means that there is no limit to how little of something you can have.
  - Ratio Scale
    - A quantitative measurement scale in which the numerals have equal intervals and the value of zero truly means “none” of the variable being measured
      - Pulse is an example of a ratio measurement. You can’t have a negative pulse – if you have no pulse, then your pulse is zero, and you can’t go below that.
  - Interval vs. Ratio
Validity and Reliability
- Validity
  - Are you accurately measuring what you claim to be measuring?
- Reliability
  - Does your measure produce consistent results?
Three Reliabilities
- Test-Retest Reliability
  - The consistency in results every time a measure is used.
  - If a construct should be stable, measures of that construct should be stable as well.
    - Weight and Height should be similar when measured on the same day; if they are not, they are not that reliable
- Interrater Reliability
  - The degree to which two or more coders or observers give consistent ratings of a set of targets.
    - Observers observing the same thing should agree on what they are seeing.
    - High and Low Interrater Reliability
    - Dealing with Poor Interrater Reliability
      - Create clear coding guidelines at the outset
      - Train coders on dummy data before coding actual data
      - Can have coders discuss differences
      - If all else fails, bring in a third coder
      - Can indicate genuinely ambiguous data
- Internal Reliability
  - In a measure that contains several items, the consistency in a pattern of answers, no matter how a question is phrased.
  - Issues in scale construction
  - Items that correlate poorly with the rest of the scale should be removed
A Deeper Dive on Construct Validity
- Construct validity = whether the measure accurately assesses the given construct
  - How can we know our measure has construct validity?
    - Construct Validity of Abstract Variables
      - Construct Validity is not All or Nothing
        Researchers talk about HOW valid a measure is
        We never talk about perfect validity
- Breaking Down Construct Validity
  - Face Validity
    - Face validity refers to the extent to which a measure seems to measure what it claims to measure.
    - Face validity does not guarantee construct validity, nor is it necessary for construct validity
    - But most measures with construct validity will also be face valid
      - Be Highly Skeptical of Measures that are Not Face Valid
    - The Myers-Briggs
  - Criterion Validity
    - The extent to which a measure is associated with a behavioral outcome with which it should be associated
    - Constructs = individual differences
    - Individuals who differ on constructs should think, feel, and act differently from one another
      - The idea of Extroversion
  - Convergent Validity
    - Convergent validity is the extent to which my scale correlates with other ways of measuring the same construct or the extent to which my scale correlates with measures of conceptually related variables.
      - One measure of convergent validity is whether scores on a scale correlate with similar constructs.
  - Discriminant Validity
    - Can the scale differentiate among related constructs?