Assessing the Consistency of Average Student Growth Using a Split Classroom Approach

Assessing the Consistency of Average Student Growth Using a Split Classroom Approach

[featured_image]
  • Version
  • Download 95
  • File Size 1.11 MB
  • File Count 1
  • Create Date August 2, 2018
  • Last Updated August 2, 2018

Assessing the Consistency of Average Student Growth Using a Split Classroom Approach

The United States Department of Education is promoting the use of student growth on summative assessments as part of teacher evaluation. As in any measurement model there are multiple sources of error variance that in this case will detract from our ability to assess the true average student growth attributable to a teacher.In this study, using data from a state testing program, the reliability of average student growth attributable to teachers was assessed in two ways: within year and across years. Within year all children are taught by a teacher in the same classroom(s) in a given year. Their current year test scores are compared with their previous year scores (when they typically had different teachers). Each class is divided into random halves and the two average student growth scores are calculated for each teacher and those scores are correlated across teachers with the same class size (class size will affect the reliability of average class growth similar to how test length affects test score reliability). We call this the split-class method. In order to calculate cross-year correlations, an average student growth score was calculated for each teacher based on all students’, growth scores of each teacher from 2010 to 2011 and then again from 2011 to 2012, and the two cross-year growth scores of all teachers are correlated within a similar range of class sizes.The sources of error variance are different with these two designs. In the first design, within each classroom there is no variation in how the class is taught. In a cross-years design the teacher or school might make changes to instruction or curriculum from one year to another. Similarly there is no differential variation in teacher health, personal circumstances, or attitudes in the split-class design, but might be in the cross-years design. In addition, the makeup of the classroom might affect a teacher’,s ability to facilitate learning from one year to the next in the cross-year design.In this study we will assess the variance due to within-year and cross-year factors using three different common growth models.

Attached Files

FileAction
paper_226dc1c022.pdfDownload