Assessment and Testing
Assessment and Testing tests whether you can select, interpret, and ethically use instruments - and whether you know what a score does and does not mean.
Core
- Reliability: test-retest (stability), internal consistency (Cronbach's alpha, split-half), inter-rater, parallel forms. Reliability is necessary but not sufficient for validity.
- Validity: content (covers the domain), construct (measures the intended trait), criterion - concurrent (now) and predictive (future). A test can be reliable and invalid.
- Standard scores: z (M=0, SD=1), T (M=50, SD=10), stanine (M=5, SD≈2), Wechsler IQ (M=100, SD=15), percentile rank (percent below).
- SEM: SD × √(1 − r); used to build confidence intervals around an observed score.
- Norm- vs. criterion-referenced: comparison to others vs. absolute standard.
- Clinical: mental status exam, DSM-5-TR criteria, suicide risk and protective factors, informed consent for testing.
Worked example
A client scores at the 84th percentile on a test with M=100, SD=15. Reasoning: percentile 84 corresponds to z = +1, so the score is about 115. It does not mean 84% of items were correct. If the reliability coefficient is .91, SEM ≈ 15 × √.09 ≈ 4.5, so the true score likely sits within a band around 115.
Traps
- Reading percentile rank as percent correct.
- Inferring validity from a high reliability coefficient.
- Confusing T scores with t tests.
- Treating grade-equivalent scores as interval data.
- Using an instrument outside its norm group without caution.