Assessment and Data-Based Decision Making
This domain covers psychometrics, assessment methods, and the problem-solving process used to make eligibility, instructional, and intervention decisions. Expect score conversions, reliability and validity interpretation, and choices about which data source answers a given question.
Score systems
- Standard score: mean 100, SD 15 (85 = −1 SD, 70 = −2 SD, 115 = +1 SD).
- Scaled score: mean 10, SD 3. T-score: mean 50, SD 10. z-score: mean 0, SD 1.
- Standard error of measurement: SEM = SD × √(1 − reliability). The 95% confidence interval is the obtained score ± 1.96 × SEM.
- Reliability types: test–retest, internal consistency, alternate form, interrater. Coefficients of .90 or higher are expected for high-stakes decisions.
- Validity types: content, construct, criterion (concurrent and predictive). Reliability is necessary but not sufficient for validity.
Assessment methods
- RIOT: Review records, Interview, Observe, Test.
- Norm-referenced tests compare to peers; criterion-referenced tests compare to a standard; curriculum-based measurement (CBM) is a repeated, rate-based general outcome measure with benchmarks and growth rates.
- IDEA requires multiple measures, technically sound instruments, assessment in all areas of suspected disability, and testing in the child's native language when feasible.
- SLD identification may use RTI/dual discrepancy, patterns of strengths and weaknesses, or ability–achievement discrepancy, depending on state rules.
Problem solving and progress monitoring
Four stages: problem identification, problem analysis, plan implementation, plan evaluation. Compare baseline, goal, aim line, and trend line; examine level, trend, variability, immediacy, and overlap. The four-point rule flags inadequate response; percentage of non-overlapping data summarizes intervention effect.
Worked example
A student earns a standard score of 85 on a measure with SEM = 4. The 95% confidence interval is 85 ± 7.8, or roughly 77–93. A later score of 92 falls inside that interval, so the two scores do not represent a reliable difference. The defensible conclusion is "no meaningful change," not "the student improved seven points."
Traps that cost marks
Reading grade equivalents as instructional levels. Basing eligibility on a single subtest or single measure. Ignoring SEM and confidence intervals when comparing scores. Treating a 15-point index difference as automatically significant.