Differential item functioning, known as DIF, is a statistical property of a test item describing how likely individuals from different groups, even when they possess comparable underlying ability, are to answer that item differently. Uniform DIF gives one group a consistent advantage across all ability levels, while nonuniform DIF means the size of the advantage changes depending on the test taker's own ability level. Detecting DIF does not by itself prove that an item is biased; it flags unexpected item behavior that then requires review and judgment. Test developers commonly identify DIF using the Mantel-Haenszel procedure, logistic regression, item response theory based methods, and confirmatory factor analysis.
Facts
Core ClaimDifferential item functioning is a statistical property of a test item showing that individuals from different groups with comparable underlying ability have different probabilities of answering it correctly, split into uniform DIF (a consistent advantage for one group) and nonuniform DIF (an advantage that varies with ability level). 1 FAQs
Why do groups have to be matched on ability when checking a test item for differential item functioning?
Without matching on the measured attribute, detection can be inaccurate.
Differential item functioning compares how an item works across groups, often demographic ones, that are matched on the attribute the test measures. When groups are not matched on that attribute, the analysis may detect differential item functioning inaccurately. The comparison is meant to show whether people with the same underlying ability have an equal chance of answering correctly, not whether the groups differ in overall score.
Sources
1. Wikipedia: Differential item functioning
Wikimedia FoundationLead paragraph, first sentenceQuote, Lead paragraph, first sentence
Differential item functioning (DIF) is a statistical property of a test item that indicates how likely it is for individuals from distinct groups, possessing similar abilities, to respond differently to the item.
Reader Challenges (0)
No disputes yet. Spotted an error or a better source? Open the first one.
Sign in to dispute this or suggest a correction.