PSYCHOMETRICS Flashcards
Define Psychometrics
measurement of theory and test applications of quantitative assessment of psychological constructs
Why are psychometric important?
- Indirect assessment and development of latent attributes
- used to make important decisions
- best, fairest, economical method
Define psychological measurement
Process of assigning numbers to a person
Limitations of psychological tests
- Precision & Accuracy
- Single psychological attribute
- assumes psychological constructs real and exist
- Psychological numbers can be represented by numbers
What is sampling?
Selecting to observe a part of target population, estimate characteristics of the population
Representativeness of sample
How representative of the sample of the population
Biasedness
Degree of systematic/random error of sample
Psychological test score
Generating single numerical value, quantitative representation
Objective scoring
standardised questionnaires
Subjective scoring
Assessor judgement
Standardisation
Process of transforming scale to universal indexes (IQ)
Reasons for Standardising
- Makes it easier to compare
2. Generations of population norms
Use of Standardising
- Records population attributes
2. Compare individual attributes to population levels.
Z-scores
z = x-u/o
What is validity?
Degree to which test measures what it claims to measure. appropriateness, usefulness, meaningfulness
Construct validity?
Degree to which constructs possess sound theoretical foundation. operationalisation
Convergence validity?
High correlation with items/tests that measure the same/related constructs
Divergent validity?
Low correlation with items/tests that assess unrelated constructs
Criterion validity?
Degree of correlation with one or more parallel or outcome criteria. correlation coefficient
concurrent validity?
criterion is present
predictive validity?
criterion is future
internal validity?
confidence in causal relationships between measured constructs
external validity?
valid construct across different settings (time, age, populations)
content validity?
scores represent the content area they claim to (e.g. content of a final should be representation of learned content)
problems with validity?
- psychological constructs are abstract and latent
- reality of psychological world – can’t be valid/invalid
- what is meaningful/useful can change
what is reliability?
degree of consistency or stability of measurement across time and context. how well does measurement assess claim?
What is classical test theory?
X = T + E (true score + error) reliability index (r) = varianceT/varianceX
Issues with CTT
- True scores – temporal score instability & chance fluctuations
- Systematic error
Individual error sources
- Idiosyncratic - language, mood, fatigue
2. Generic - faking, random responses
Item error sources
- content related - lack of clarity
- format related - biases
- admin related - distracting settings
internal consistency reliability
degree of consistency in response to scale items of same construct. cronbach alpha
cronbach alpha cut-offs
conservative 0.8 liberal 0.6
test-retest reliability
assess stability of test scores over time. usually take the test at least 2 diff times
issues with test-retest
- drop-outs/non-response bias
- temporal instability of constructs
- optimal time-interval
inter-rater reliability
reliability based on 2+ independent judges who rate the items
standard error of measurement
average degree of random error
what is confidence interval
CI is range of observed scores within which the true score is expected to lie
What is factor analysis?
FA is referred to advanced multivariate statistics which help uncover latent structure/construct from a set of observed attributes
FA used for?
- condense large number into smaller latent constructs
2. holistic assessment tool
what is item in FA
observed measured element of attribute
what is factor/component in FA
latent dimension (construct) made up of related items
Factor loading?
Statistical relationship between factor and item. > |0.4| is considered good
When is factor considered potentially meaningful?
the items loaded need to be related quantitatively and qualitatively
major FA types (3)
- CFA (confirm or validate already hypothesised construct)
- EFA (exploratory, identify possible construct)
- PCA (principal components; assumes all variance can be explained by same latent structure)
What is orthogonal factors
dimensions independent from each other e.g. neuroticism and extraversion
what is oblique factors
dimensions related to each other e.g. Gf and Gc factors of intelligence
factorability?
suitability of item to be included in FA model
simple structure?
items form distinct factors or spatial clustered, based on degree of associations
rotation?
geometric transformation to generate model that contains simpler structure
varimax rotation?
orthogonal rotation - rotate unrelated, maximises the variance
oblimin rotation?
oblique rotation, more related factors
Kaiser criterion?
retain any factor with eigenvalue of greater or equal to 1
Scree plot rule
variance explained rule (retain all 80-90% total variance) comprehensibility rule (retain all that are clearly interpretable
applications for psychometric knowledge
- criminal/clinical profiling
2. psychobiography/psychohistory