ERIC Number: ED202916
Record Type: RIE
Publication Date: 1981-Apr
Reference Count: 0
Will Removing a Few Bad Apples Save the Barrel?
Doss, David A.
For many reasons, students do not always perform on tests in ways that are congruent with their true abilities or achievement levels. Problems of this sort may be especially common on Title I evaluations where the tests administered are too difficult for a large percentage of the students. Such considerations raise the questions of whether anything can be done to identify and remove scores which appear to be "invalid" so that some meaning can be gleaned from the evaluation. The effects of using the Rasch person-fit statistic to remove students with possibly invalid scores from a Title I Model C analysis were investigated. First a Model C analysis was computed using all students. Then those students with a reading subtest fit statistic in the top 10% district-wide were removed and the Model C analysis was redone. Then students with scores in the top 20% and the top 30% were removed. The results showed that removing some students produced a modest change in the evaluation outcome and that removing more students did not lead to significantly more change. The results support the value of using the Rasch person-fit statistic to identify students who do not fit the model. The consistent reduction of the standard error of estimate as more students were removed seems to indicate that the procedure identifies the "right" students. (Author/GK)
Publication Type: Speeches/Meeting Papers; Reports - Research; Reports - Evaluative
Education Level: N/A
Authoring Institution: N/A
Identifiers: California Achievement Tests; Iowa Tests of Basic Skills; One Parameter Model; Rasch Model; Title I Evaluation and Reporting System
Note: Paper presented at the Annual Meetinq of the American Educational Research Association (65th, Los Angeles, CA, April 13-17, 1981).