Applied Measurement with jMetrik by J. Patrick Meyer

Author:J. Patrick Meyer [Meyer, J. Patrick]
Language: eng
Format: azw3, pdf
ISBN: 9781136294167
Publisher: Taylor and Francis
Published: 2014-06-13T04:00:00+00:00

6

Differential Item Functioning

Differential item functioning (DIF) occurs when one group of examinees has a different expected item score than comparable examinees from another group. It indicates that an item is measuring something beyond the intended construct and is contributing to construct irrelevant variance. It also suggests that the item is biased and unfair for some examinees. DIF is a threat to validity that should be identified and eliminated to improve the quality of inferences about the measured trait.

A DIF analysis compares the performance of two different groups of examinees. The focal group is the one of interest, and it usually represents a minority group such as black students or English language learners. Their performance is compared to a reference group that represents a larger group of examinees such as white students or native English speakers. A population of examinees may contain multiple focal and references groups, but a single DIF analysis involves only one pair of groups (e.g. Asian and white examinees). The analysis is repeated with different pairs of groups to evaluate DIF for a variety of focal and references groups.

One way to evaluate the performance of focal and reference group members is to compute the difference in mean item scores. This difference is referred to as item impact (Clauser & Mazor, 1998; Holland & Thayer, 1988), and it reveals a difference between the two groups, but it does not indicate a problem with the item. Item impact could be due to group differences on the measured trait. DIF is a distinct concept from item impact. DIF procedures control for group differences on the measured trait by evaluating the performance of comparable focal and reference group members. Differences in item performance for comparable focal and reference group members indicate a problem with the item that warrants further review.

Figure 6.1 illustrates DIF using two item characteristic curves. The curve on the left describes the probability of a correct response for male examinees, and the curve on the right shows the probability of a correct response for female examinees. DIF is indicated in this graph because for a given value of the measured trait (x-axis), males have a higher probability of a correct response (y-axis) than females. For example, at a value of Theta = 0, females have about a 25% chance of a correct response, whereas males have about a 72% chance of a correct response. Moreover, males have a greater probability of a correct response at all other values of the measured trait. This consistent advantage for one group is referred to as uniform DIF.

FIGURE 6.1 Plot of uniform DIF

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.