EGATSU-PI flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

Scaled score

A transformed raw score placed on a common reporting scale so that scores from different test forms or editions can be compared and interpreted against a fixed reference.

Raw score

The number of items a candidate answers correctly before any weighting, scaling, or adjustment; it is form-dependent and cannot be compared directly across different test forms.

Cut score

The minimum score, set by a standard-setting panel, that separates passing from failing performance; it is a policy decision, not a statistical property of the items.

Angoff method

A standard-setting procedure in which judges estimate, for each item, the probability that a minimally competent candidate answers it correctly; the sum of those probabilities becomes the recommended cut score.

Standard error of measurement

An index, computed as the standard deviation multiplied by the square root of one minus reliability, that estimates how much an observed score is likely to vary around a candidate's true score.

Reliability coefficient

A statistic between zero and one describing how consistently a test ranks candidates across repeated measurements; higher values mean less measurement error and narrower confidence intervals around reported scores.

Item difficulty p-value

The proportion of examinees answering an item correctly, ranging from 0 to 1; on a four-option multiple-choice item, values near 0.63 typically maximize discrimination.

Item discrimination index

A statistic, often the point-biserial correlation or the difference between upper- and lower-group pass rates, showing how strongly an item separates high-scoring from low-scoring candidates.

Distractor analysis

Reviewing how often each incorrect option was chosen; a distractor selected by no one may be implausible, while one chosen more often than the key signals a flawed item.

Criterion-referenced test

An assessment interpreted against a fixed performance standard rather than against other examinees, so every candidate can theoretically pass or fail regardless of the group's ability distribution.

Norm-referenced test

An assessment interpreted by comparing a candidate's score with the scores of a defined reference group, typically reported as percentile ranks or standard scores such as z or T.

Item response theory

A family of models that expresses the probability of a correct response as a function of latent ability and item parameters such as difficulty, discrimination, and guessing.

Rasch model

A one-parameter item response theory model in which items differ only in difficulty and the log-odds of success equal the difference between person ability and item difficulty.

Guessing correction

A formula-score adjustment that subtracts a fraction of wrong answers, usually one divided by the number of options minus one, intended to remove the advantage of random responding.

Test blueprint

A specification table that maps content areas and cognitive levels to the number of items, ensuring the assembled form matches intended coverage and weighting.

Content validity

Evidence that a test's items adequately represent the defined content domain, usually established through expert review against a blueprint rather than through correlation with other tests.

Test-retest reliability

A stability estimate obtained by administering the same test to the same group twice and correlating the two sets of scores; suitable only when the trait is expected to remain stable.

Cronbach's alpha

An internal-consistency estimate derived from the number of items, their variances, and the total test variance; it approximates the average correlation among all possible split-halves.

KR-20

The Kuder-Richardson formula 20 gives internal consistency for dichotomously scored items; it equals Cronbach's alpha when every item is scored right or wrong.

Equating

A statistical process that places scores from different test forms onto a common scale so that a given scaled score implies the same proficiency regardless of which form was taken.

Percentile rank

The percentage of a reference group scoring at or below a given score; percentile ranks are ordinal and cannot be averaged or arithmetically combined.

z-score

A standard score expressing how many standard deviations a value lies above or below the mean, computed as the raw score minus the mean, divided by the standard deviation.

T-score

A standard score with a mean of 50 and a standard deviation of 10, obtained by multiplying the z-score by 10 and then adding 50.

Confidence interval

A range around an observed score, built from the standard error of measurement and a chosen confidence level, within which the true score is likely to fall.

Differential item functioning

A condition in which candidates of equal ability but different group membership have different probabilities of answering an item correctly, indicating possible item bias.

Item exposure control

Procedures in computer-based testing that limit how often each item is administered, protecting bank security while preserving measurement precision across candidates.

Computer-adaptive testing

A delivery mode that selects each item based on the candidate's running ability estimate, shortening the test while maintaining precision near the reported score.

Passing standard

The defined level of knowledge or skill required to pass, translated into a cut score on the reporting scale through standard-setting and subsequent equating.

Time-per-item budget

A pacing rule computed by dividing available testing minutes by the number of items, giving the average time a candidate can spend before falling behind.

Standard-setting panel

A group of qualified judges who review items and performance data to recommend a cut score, documenting the rationale behind each judgment they make.

Item bank

A secure, calibrated collection of items tagged by content, difficulty, and exposure history from which test forms or adaptive selections are assembled.

Anchor items

Items shared between two test forms that link them statistically; their difficulty differences are used to place both forms on a common scale.