Comparing Apples with Apples: Assessing Student Attitudes in the Presence of Regression to the Mean

• Activity Number: 342
• Type: Topic Contributed
• Date/Time: Tuesday, August 6, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Education
• Author(s): Anne Michele Millar*+ and Bethany White
• Companies: Mount Saint Vincent University and Western University
• Keywords: statistics education ; regression-to-the-mean ; gain scores ; measurement error ; attitudes

Abstract

When we assess students' attitudes towards statistics we are usually more interested in how their attitudes have changed during the course, than their final attitudeds. Typically, the gain in attitude (post-score -- pre-score) depends on the pre-score: negative attitudes tend to have a higher gain score, while more positive attitudes tend to have a lower one. When we compare results across groups of students, we need to consider students gains relative to their pre-scores for a meaningful comparison - apples with apples. Since attitude scores include measurement error, the "true" relationship between the gain and the pre-score is difficult to estimate due to regression-to-the-mean. We compare methods for parameter estimation and confidence interval construction to accurately estimate gains in attitude relative to pre-scores, allowing for measurement error.