Psychometrics

1
Psychometrics

• Evaluating Objective Measure of Personality
  Traits

2
Reliability

• Does measure yield consistent scores over time,
  observers, and items
• Estimates error component
• Three types
• Temporal
• Inter-rater
• Internal Consistency

3
Temporal Reliability

• Does measure yield consistent scores over time
• Assess Test-Retest reliability coefficient
• Measure trait at Time 1 and Time 1(less than 6
  months) and compute correlation

4
Inter-rater reliability

• Does measure yield consistent scores over
  observors
• Required for measures with subjective component
  (e.g., projective, behavior)
• Assess with agreement rate
• Low agreement rate indicates high error
  (idiosyncratic observer perceptions)

5
Internal Consistency

• Consistency over items
• Increasing number of items increases temporal
  reliability However, items need to measure
  same trait
• Assess Split-half reliability coefficient
• Correlate scores of two halves (randomly
  selected)

6
Internal Consistency continued

• Average item-total correlation
• Average correlation of each item with total
• Example Extroversion
• Item Person 1 Person 2 Person 3
• friends 30 5 20
• phone calls 35 2 4
• football games 6 5 10
• parties 15 1 10
• Total 80 13 44

7
Validity

• Does measure tap intended trait?
• Reliability necessary but not sufficient for
  validity

measure
Trait
Measure
Any trait
8
Validity Types

• Construct Validity
• Theoretical what is measure assessing? How
  does it relate to other constructs?
• Criterion Validity
• Practical does measure serve intended purpose?

9
Construct Validity

• Convergent. Measure should correlate with other
  measures of same/similar construct.
• Correlate using different assessment types
• E.g., Extroversion with peer reports, behavior
• Discriminant. Measure should not correlate with
  other measures of different constructs
• E.g., Extroversion with social anxiety, social
  desirability

10
Criterion Validity

• Predictive
• Does measure predict future behavior related to
  measured trait?
• E.g., GRE, SAT, job placement
• Concurrent
• Does measure differentiate between groups known
  to differ?
• E.g., Depression measure should discriminate
  between (diagnosed) depressed and controls

11
(No Transcript)