Abstract
Assumptions and considerations for detecting differential item functioning in empirical data. This paper examined three of the most common assumptions in DIF analysis, and the consequences of their violation. First, the samples used for DIF analysis are statistically representative samples of the specified populations. Second, the vast majority of items in the test are valid items. Third, the cost of making a type I error is greater than the cost of making a type II error. As well, the implications that for empirical DIF studies show the statistical and practical significance, the characteristics of items and populations, and finally, the coherence between complementary tecniques, were examined. In short, some cautions are offered regarding the use of statistical DIF analysis with empirical data.
