Assessment and Testing
What gets measured
Assessment items test psychometric concepts first, instruments second.
Reliability (consistency) covers internal consistency (coefficient alpha, split-half), test-retest, parallel forms, and interrater agreement. Validity (does it measure what it claims) covers content, criterion (concurrent, predictive), construct, and face validity - face validity is not evidence of accuracy. A test can be highly reliable and still invalid.
Standard error of measurement: SEM = SD x square root of (1 - r). It defines the score band around an obtained score. Scores are usually reported as standard scores: z (mean 0, SD 1), T (mean 50, SD 10), IQ/standard score (mean 100, SD 15), stanines (1-9), and percentiles (rank within the norm group, not percent correct).
Common instruments: WAIS and WISC (intelligence), Stanford-Binet, Woodcock-Johnson; MMPI-2/MMPI-3 and NEO-PI (personality); Beck Depression Inventory, PHQ-9, GAD-7 (symptom screens); Strong Interest Inventory and Self-Directed Search (interests); Conners and BASC (behavior ratings); projectives such as the Rorschach and TAT, which raise reliability and fairness concerns.
Ethics: know the DSM's definition of a mental disorder, use suicide risk tools (C-SSRS, ASQ) as adjuncts to clinical interview, secure test data, release results only to qualified professionals, obtain informed consent, and honor ADA/504 accommodation requirements.
Traps: calling the SEM a validity index; treating a percentile as percent correct; assuming a high screening score is a diagnosis; using grade-equivalent scores to compare growth.
Rule of thumb: never report an obtained score as a point - report it as a band of roughly plus or minus one SEM.