Diskriminationsindex

Diskriminationsindex, in the field of psychometrics often abbreviated as DI, is a statistic that measures the extent to which a test item differentiates between examinees with high versus low overall ability. It is used to assess item quality in tests and questionnaires.

Several variants exist. The most common is the top–bottom discrimination index: compute the proportion of correct responses among the high-ability group (often the top 27% by total test score) and subtract the proportion of correct responses in the low-ability group (bottom 27%). A higher DI indicates better discrimination. Other widely used measures include the point-biserial correlation between the item score (0 or 1) and the total test score, and the corrected item-total correlation (rit) obtained by correlating item score with total score excluding the item.

Interpretation: For correlation-based measures, values range from -1 to +1, with higher positive values indicating stronger discrimination. For the difference-of-proportions version, the index also lies between -1 and +1; values near 1 indicate good discrimination, near 0 indicate little discrimination, and negative values may indicate miskeyed or problematic items.

Applications: used during item analysis in test development to guide item revision or removal and to provide overall feedback on test quality. It is especially relevant for multiple-choice tests but can be used for dichotomous items in general.

Limitations: the index depends on the sample and the test’s overall difficulty and dimensionality; it can be biased by guessing; small samples yield unstable estimates; when tests are multidimensional or targeted to subgroups, differential item functioning should be examined.

Related concepts include item difficulty, item-total correlation, and differential item functioning.