Assumptions, Cautions, and Solutions in the Use of Omitted Test Data to Evaluate the Achievement of Underrepresented Groups in Science-Implications for Long-Term Evaluation

Often science test results are used to evaluate differences in achievement between males and females. These test scores and subsequent calculations are commonly utilized to guide and monitor reform efforts. Analysis of data collected at the statewide level in Ohio suggests that it can be important to consider the rate of item omission before scores are computed. In nonreform classrooms a great difference in test item omission rates was observed between African Americans and whites as well as males and females. In reform classrooms very little difference in omission rates was observed. This mixture of patterns suggests that evaluation of students must also include a careful investigation of how much of a test was not answered by individuals. If test items omitted are counted as incorrect, the achievement level calculated for students may be underestimated. Also, if items within subscales are not reached by particular groups of test takers, then performance with regard to one specific topic may be underestimated. Although omitted data can influence calculations, a variety of steps can be taken to consider data, and statistical techniques can be utilized that do not count omitted items as incorrect. By monitoring missing data, program evaluation of equity activities can measure achievement and not just an ability to quickly respond to testing instruments.