ERIC Number: ED406187
Record Type: RIE
Publication Date: 1997-Mar-22
Race, Gender, Test Length, and Missing Data: Why Estimates of Performance May Be Clouded.
Boone, William J.; And Others
In 1991, Ohio received National Science Foundation (NSF) funding through its Statewide Systemic Initiative (SSI) program. One aspect of the reform effort involved evaluating the performance of middle school students with a test item bank of items from the National Assessment of Educational Progress (NAEP). This paper presents the results of evaluating these data. It explores how unanswered items can effect analysis of such data when it is used to calculate mean performance measures of groups. How "missing" data can influence calculations of group performance is significant, for if particular subgroups do not complete a test in much higher numbers than other subgroups, it is likely that analyzed data may not reflect reality. Analyzed data showed a great disparity in the percentage of blacks and whites answering the science test items. Noteworthy are black and white students' answering patterns toward the end of the science test. Findings indicate that male and female test takers exhibit some of the same trends as observed in the racial composition. It is concluded that the design of science tests can greatly influence the quality of achievement measures calculated for students. When tests are administered to students, it is critical to evaluate the influence missing data may have upon calculations. (Author/JRH)
Descriptors: Academic Achievement, Data Analysis, Educational Change, Intermediate Grades, Junior High Schools, Middle School Students, Middle Schools, Minority Groups, Racial Differences, Sex Differences, Test Results
Paper presented at the Annual Meeting of the National Association for Research in Science Teaching (Chicago, IL, March 22, 1997).
Publication Type: Reports - Research; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A