Psychometrics
The record behind the pass mark.
A cut score is not a number somebody typed on a program. It is the end of an argument, and when that argument is challenged the body has to produce it. Qualeron keeps the argument — and the analyses, the equating and the report that stand behind it.
01 Scoring and the cut score
The record behind the pass mark
- Scaled score against a passing scaled score, or pass/fail — declared per program, exclusive at the database level
- The cut score locks the moment a candidate registers
- The way through is a standard-setting study, adopted: method, date, facilitator, panel, rounds, recommendation, rationale
- A superseded study is marked, never deleted
A program scores either on a scaled score against a passing scaled score, or pass/fail. The two are mutually exclusive at the database level: a scaled program must have a passing score and a pass/fail program must not have one, so a program cannot exist in a state where nobody can say what passing means. Scaled scores are derived rather than stored as a raw count. They can support a common reporting scale across forms of different difficulty when they are backed by an appropriate, documented linking or equating method — which is why Qualeron keeps the approved equating study behind every equated conversion it applies.
A program's passing score can be changed freely while nobody has registered. Once a single candidate holds a seat, it locks. Moving a cut score reclassifies results that have already been recorded, and a system that allows it quietly is a system whose results cannot be defended. The way through is not an override. It is adopting a standard-setting study, which changes the number and records why.
A study records where the number came from: the method used, the date it was conducted, who facilitated it, who sat on the panel, the definition of a minimally competent candidate the panel rated against, the rounds, the recommended cut and the rationale. It is drafted, then adopted — adoption being the act that makes it the program's standard. A body recording a standard it set years before it came to Qualeron can adopt a study that documents the cut already in force, without disturbing it. Panel names and ratings are your records, visible to your people and to nobody else.
02 The analyses
Governed, reproducible, independently approved
- Every analysis runs over a frozen population with a versioned algorithm
- Draft → submitted → approved by a different person, or returned; a successor supersedes
- An analysis never changes a score, a form, a cut score or a result
- The technical report's evidence manifest is hashed and frozen
A psychometric analysis is a reproducible calculation over a frozen population, using versioned statistics, wrapped in a governed lifecycle the database enforces: run, draft (or insufficient data), submitted for review, approved by someone other than the person who ran it — or returned, or withdrawn. It never changes a score, a form, a cut score or a result, and it never calls an estimate good or bad. It reports what was computed, from which sittings, by which algorithm version, and who reviewed it. Interpretation belongs to a qualified person, who writes it down.
- Reliability and SEMCoefficient alpha, KR-20 and the standard error of measurement for an administered form.
- Decision consistencyHow consistently the form classifies candidates against the cut score.
- Fairness reviewMantel–Haenszel differential item functioning with purpose-limited groupings, so a review of fairness cannot become a profile of the people who sat.
- EquatingCommon-item nonequivalent-groups studies over designated anchors — chained linear, with chained mean as the sensitivity check — applied to scoring only through a governed decision that records why.
- The technical reportOne administered form's evidence in one place: the form and its blueprint, volume, item statistics, the latest approved analyses, the adopted standard-setting study, equating studies and applied conversions, incidents, challenge decisions and score changes — compiled once into a manifest that is hashed and frozen, with missing evidence recorded as missing, never as a zero. People write the narrative.
03 Score reporting
By content area, if you want it
- Per program, off by default
- Bands by content area, not a bare pass or fail
- Release, holds, challenges, reviews and appeals on the record
A program can report a candidate's performance by content area in bands rather than as a bare pass or fail, so a candidate who did not pass learns something about where they stand. It is per-program and off by default, because a domain report on a form with too few items per area is a number that looks informative and is not.
Results are released on your schedule, held when someone should look first, challenged item by item within a window you set, reviewed and appealed with a reason a candidate reads — the due process described on the platform page.
04 Practice analysis
The study behind the blueprint
A job or task analysis is the argument for the blueprint the way a standard-setting study is the argument for the cut score, and Qualeron keeps it the same way. A study records its design and its expert panel with the composition it was meant to have; the outline of domains, tasks and knowledge is revised with its lineage intact; a validation survey goes out from a frozen snapshot of the outline by one-time links, and responses are pseudonymous. Results are computed under a rule declared before the survey opened; a decision to depart from it is recorded with its reason, and the respondents are compared with the benchmarks your body set for who should have answered.
- A study design and a documented expert panel with composition targets
- A domain, task and knowledge outline with revision lineage
- A validation survey from a frozen snapshot, one-time links, pseudonymous responses
- Results under a predeclared rule; departures decided with reasons; representativeness against your benchmarks
Related
- Item statistics & qualityDifficulty, discrimination, distractor analysis and miskey alarms on every form.More about Item statistics & quality
- Blueprints & assemblyContent outlines with constraints, constrained form assembly, anchors, exposure controls and form pools.More about Blueprints & assembly
- Accreditation evidenceThe records an NCCA or ISO/IEC 17024 submission asks for, produced as a by-product of running the program.More about Accreditation evidence
- Roles & permissionsOwners, administrators, examiners, reviewers and psychometricians with the access each role needs, and the audit trail behind changes.More about Roles & permissions
Ask what your program would need
Tell us about your program.
Tell us how many candidates you certify and what you run today. You will get a written quote against the published plans — not a discovery call before a number.