Resumen

The detection of differential item functioning (DIF) is fundamental to ensure instruments’ invariance, and, therefore, a better estimate of the construct being measured across the different groups of examinees. The purpose of this research was to provide substantive hypothesizes related to possible sources of DIF, comparing students reported with accommodations for ADHD (focal group) and students with no accommodations (reference group), using the Standardized-P Difference and the Mantel Haenszel (MH) methods. Data from the Costa Rican national high school exit tests in Spanish and Math, from the year 2004, in public schools, were analyzed. First, these two methods were used to detect items with DIF, and then, using a more qualitative approach, drew hypotheses related to possible sources of DIF. Some degree of agreement was verified between the two different empirical methods, being Mantel-Haenszel more sensitive. In the Spanish test, DIF was hypothesized to be caused by the length and wording of the options, and the literary and non-literary texts in the stem. In Math, possible sources of DIF involved vocabulary, wording, the transition of verbal to mathematical language, the visuospatial item organization, and the drawing of graphs and geometrical figures. At the methodological level, complementing the statistical analyses with the judges’ criteria was helpful to identify possible sources of irrelevant variance in the construct measured by these tests. The hypotheses must be interpreted with caution, though, since the number of items detected as exhibiting DIF was relatively small.