EPPP Assessment and Diagnosis: Reliability, Validity, Base Rates and the Score Math the 16% Area Tests

What the EPPP assessment and diagnosis area covers

Psychometrics, instruments and diagnosis share one content area on EPPP Part 1, and it's a heavy one: ASPPB's current content areas give Assessment and diagnosis 16%, about 28 of the 175 scored items, through 13 knowledge statements (KN29 to KN41). They fall into four groups:

ASPPB's 2026 doctoral program report puts the 2021–2025 average for accredited-program candidates at 68.89% correct, third-lowest of the eight areas. Combined with the heavy weight, that makes it the area where the average candidate loses the most scored items, about 9, as the EPPP difficulty guide works out. In Passy's view, the psychometric half is the most mechanical part of this area, so it rewards drilling.

Reliability on the EPPP: three coefficients and scorer consistency

The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014), which ASPPB names in its ethics outline (KN62), recognize three broad categories of reliability coefficients, defined by how the data were collected, plus indices of scorer consistency where scoring involves judgment:

Two practical reads of this table. First, match the coefficient to the question: the Standards point out that a coefficient may reflect only the internal consistency of item responses and miss the error from day-to-day changes in performance, so a high alpha says nothing about stability over time. Second, the Standards caution that when raters in everyday settings are less well trained than those in a reliability study, the error from rater variability can be much higher than the reported interrater coefficients suggest.

The standard error of measurement and score bands

The standard error of measurement (SEM) turns reliability into score points. As a 2006 article in Health and Quality of Life Outcomes puts the formula:

SEM = SD × √(1 − reliability)

Work it through on a scale with a mean of 100 and an SD of 15 and a reliability of .91:

1. 1 − .91 = .09, and √.09 = .30.
2. SEM = 15 × .30 = 4.5 points.
3. A 95% band uses about ±1.96 SEM, the same 1.96 that article uses for a 95% interval: 1.96 × 4.5 ≈ 8.8.
4. An obtained score of 100 sits in a band of about 91 to 109.

Two patterns are worth memorizing. Higher reliability means a smaller SEM, and a perfectly reliable test (1.00) would have an SEM of 0. And at a reliability of .75, the SEM is exactly half an SD, a point the same article makes. When an item asks for the range "68% of the time," use ±1 SEM; for 95%, use about ±2.

Validity: the sources of evidence the Standards describe

The Standards define validity as the degree to which evidence and theory support the interpretations of test scores for proposed uses. In their words, it's the interpretations of test scores for proposed uses that are evaluated, not the test itself. They organize the evidence into sources:

• Test content. How well the content matches the construct.
• Response processes. Whether test takers use the cognitive processes the interpretation assumes.
• Internal structure. Whether the relationships among items fit the construct. The Standards' glossary describes internal structure as the factorial structure of item responses or subscales.
• Relations to other variables. Convergent evidence (relationships with measures of the same or similar constructs), discriminant evidence (relationships with measures of different constructs), and test-criterion relationships.
• Consequences of testing. Evidence about the consequences of test use, including whether unintended consequences arise from the test measuring something other than the intended construct, or missing part of it.

The Standards keep two classic designs for test-criterion evidence: a predictive study relates scores to criterion data collected later, and a concurrent study collects both at about the same time. They also name two threats to valid interpretation: construct underrepresentation, when a test misses important aspects of the construct, and construct-irrelevant variance, when scores reflect something else, such as a response bias to underreport anxiety on an anxiety measure.

Sensitivity, specificity and base rates on the EPPP

KN29 names sensitivity and specificity, and KN37 and KN38 name base rates. A 2017 review in Frontiers in Public Health defines them from a two-by-two table of test result against true status (a = true positives, b = false positives, c = false negatives, d = true negatives):

The review stresses that these values aren't fixed properties of a test: they depend on how stringent the cutoff is and on the prevalence of the condition in the sample. Predictive values are where base rates bite. Here's Passy's worked example, with a test that has 90% sensitivity and 90% specificity, applied to 1,000 people:

The same test is right about a positive result less than a third of the time in the low-base-rate setting. When an item gives you sensitivity, specificity and a base rate, build the 1,000-person table: it takes under a minute and avoids formula mistakes. Full EPPP prep planned for launch has explanations planned to work SEM and base-rate items step by step, the kind of calculation this section covers.

Standard scores and percentiles

Norm-referenced scores all express distance from the mean in SD units:

• z-score: (score − mean) ÷ SD, so the mean is 0 and the SD is 1.
• T-score: mean 50, SD 10, the metric PROMIS measures report, as a 2022 study in Quality of Life Research describes.
• Deviation IQ-style scores: mean 100, SD 15, the scale of measures such as the WISC-III, as noted in a 2008 article in the Journal of Neurodevelopmental Disorders.

So z = +1 equals T = 60 and a standard score of 115, and z = −2 equals T = 30 and 70. On a normal curve, about 68.3% of scores fall within 1 SD of the mean and about 95.4% within 2 SD. That puts z = +1 near the 84th percentile, z = −1 near the 16th and z = −2 near the 2nd. Those percentages are properties of the normal distribution, computed by Passy. The Standards' glossary defines a percentile rank as the rank of a score based on the percentage of scores in a specified distribution that fall below it, which is why a percentile only means something relative to the norm group.

Diagnosis and classification in ASPPB's assessment outline

KN33, KN36, KN38 and KN39 ask about diagnosis as a process: integrating non-psychological information, the rationale and limits of classification systems, dimensional versus categorical approaches, epidemiology and base rates, and models of psychopathology. Two classification systems frame it:

• DSM-5-TR. The American Psychiatric Association's DSM page describes the text revision as adding a new diagnosis, prolonged grief disorder, clarifying criteria for more than 70 disorders, and adding ICD-10-CM symptom codes for suicidal behavior and nonsuicidal self-injury. It also notes a review of the impact of racism and discrimination on diagnosis.
• ICD. The World Health Organization's ICD page says ICD-11 was adopted by the World Health Assembly in 2019 and came into effect on January 1, 2022. DSM-5-TR diagnoses carry ICD-10-CM codes, and the DSM page says no ICD-10-CM code changes are planned for DSM-5-TR in 2026.

Study diagnosis from the manuals themselves, not from summaries. For the exam, the reasoning matters as much as the labels: KN33 asks you to integrate non-psychological information into an assessment, KN38 to weigh how common a disorder is, and KN37 to recognize how base rates, group differences, cultural biases and heuristics shape interpretation.

Testing ethics inside the assessment area

Assessment topics have an ethics side. The APA Ethics Code separates test data (scores, the client's responses and your notes, released to people named in a client's release under Standard 9.04) from test materials (manuals, protocols and questions, kept secure under 9.11). The EPPP ethics guide works through that distinction with a practice item.

Two Passy-written assessment items

These items were written by Passy for practice. They aren't ASPPB items.

Item 1. A scale reports T-scores (mean 50, SD 10) and has a reliability of .84. A client obtains a T-score of 60. Which range is the client's true score most likely to fall within, at about 95% confidence?

• A. About 52 to 68
• B. About 56 to 64
• C. About 58 to 62
• D. About 40 to 80

Keyed answer: A. SEM = 10 × √(1 − .84) = 10 × .40 = 4. A 95% band is about ±1.96 SEM, or ±7.8 points, so 60 ± 8 gives about 52 to 68. B is ±1 SEM, the 68% band. C is too narrow for any conventional band. D uses 2 SD of the scale instead of the SEM.

Item 2. A screening test for a disorder has 80% sensitivity and 90% specificity. In a clinic where 10% of patients have the disorder, about what share of patients who screen positive actually have it?

• A. About 47%
• B. About 80%
• C. About 90%
• D. About 10%

Keyed answer: A. In 1,000 patients, 100 have the disorder and 80 of them screen positive. Of the 900 who don't, 10% (90) screen positive. PPV = 80 ÷ (80 + 90) ≈ 47%. B is the sensitivity and C the specificity, not predictive values. D is the base rate, which ignores the test result.

If Item 2 took you longer than a minute, build the table a few more times. The planned assessment and diagnosis question set, 216 items, is where full EPPP prep will drill this area once it launches.

Studying EPPP assessment and diagnosis efficiently

• Drill the four calculations until they're automatic: SEM bands, the 1,000-person predictive-value table, z-T-standard score conversions, and percentiles at ±1 and ±2 SD.
• Learn the vocabulary in pairs: sensitivity vs specificity, PPV vs NPV, convergent vs discriminant, predictive vs concurrent, underrepresentation vs irrelevant variance.
• Pair this area with research methods. Reliability, validity and statistics overlap, and the research methods and statistics guide covers power, effect size and design.
• Give it its full share of time. The EPPP study plan suggests about 17% of content time, a little above its weight.

Start with the 20 free EPPP practice questions to see where you stand. For the recall side of this area, 93 assessment flashcards planned for full prep cover terms such as these between longer sessions; checkout isn't open yet.