Psychometrics Flashcards
What is psychometrics?
Science of measurements
How to:
Quantify measurement tools
Validate measurement tools
Score measurement tools
Difference between assessment and test?
Assessment - skilled and trained person eg psychologist
Test - untrained to administer the test
You can fail a test not an assessment
What do psychometric tests measure?
Personality Ability/intelligence Motivation and values Clinical constructs (depression etc) Economics
What are the big five traits?
Openness Contentiousness Extravertism Agreeableness Neuroticism
What traits do psychometrics measure?
Diagnosis Job selection Job fit Understanding personality Model relationships Ability
Types of latent construct?
Personality
Intelligence
Experiential workload
5 steps to develop psychometric test
Identify possible constructs/facets
Item generation
Pilot test, analysis and review
Full test, clean data, EFA on review
Complete CFA, check for indices until good for or model is abandoned
3 ways to test reliability
Test/retest
Parallel form
Split half
What is the test retest approach?
Gold standard
People need to be tested twice so time consuming and might skew results
What is the parallel form approach?
Hardest approach
Practically simpler, but need two different versions of the same measure being tested.
Need enough participants to do order of tests in reverse
Poor test environment- any influence eg building work, affects both groups
Costly to validate an absolute and parallel test
What is the reliability split half approach?
Where components of the construct are tested with different groups
Avoids fatigue
Simple to set up, less expensive than test retest
Shorter measures are less reliable than longer measures (due to internal consistency)
Use spearman brown formula to estimate true reliability
How to maintain reliability and reduce error in testing?
Ensure assessors are familiar with instructions
Standardise test conditions
Give participants same and full instructions
4 types of validity?
Predicative
concurrent
Construct
Criterion
Explain the predictive validity
Gold standard
Does the measure predict the outcome? Eg pre employment test predict performance
Issues with predictive validity
Must bear costs of testing all applicants, even unsuccessful ones
Removal of unsuitable candidates results in bias reporting of results
People need to be in role long enough to see performance meaning lag between collection of the prediction and results of the real world implementation
What is concurrent validity?
Whether a current test correlates with a new test
Issues with concurrent validity
Motivation issues - change of behaviour to appear more desirable
Restriction of range - poor performers do not get role, top performers move on
Self awareness of personality tests might mean people try to predict what people are looking for
What is construct validity?
Are we measuring the correct construct
Unidimensionality
Convergent validity
Discriminant validity
Nomological validity
What is a unidimensional construct?
Something that measures only one construct eg depression or fatigue, not both
What is convergent validity
That two construct measures converge?
What is divergent validity?
Do two measures diverge if they should be measuring different things
Nomological validity?
Degree to construct is supported by other theories
