Combating Ceiling Effects: Modeling High-Ability Student Growth Using Multilevel Tobit Regression

• In Event: Assessment in Gifted Education

Abstract

Pressures from accountability testing have resulted in narrowing of the focus of standardized achievement tests resulting in an inability to capture the achievement or growth of high ability students. This study proposed that the Tobit model could be used to more accurately model high ability student growth to lessen the pressures on teachers and create an environment better suited for high ability student learning. Ultimately, Tobit models using artificially censored data were able to come close to replicating uncensored growth estimates under certain conditions. The results indicated that Tobit regression was necessary when examining homogeneous groups of high ability students. Finally, the Tobit regression models were able to increase the growth estimates for high ability students using naturally censored data.

Authors

• Julia Hujar, Southern Methodist University
  Presenting Author

• Richard G. Lambert, University of North Carolina - Charlotte
  Presenting Author

• Michael S. Matthews, University of North Carolina - Charlotte
  Presenting Author

• Kyle T. Cox, University of North Carolina - Charlotte
  Presenting Author

• Stella Yun Kim, University of North Carolina - Charlotte
  Presenting Author