Documenting Reading Achievement and Growth for Students Taking Alternate Assessments

Abstract

Students with disabilities have been included in state accountability systems for more than a decade; however, only in the past few years have alternate assessments of alternate achievement standards (AA-AAS) become stable enough to allow examination of these students' achievement growth. Using data from Oregon's AA-AAS in Reading during the period , we examined the achievement growth for a sample of 1,061 elementary students using two growth models: a transition matrix and a multilevel linear growth model. The authors found with the transition matrix model that a majority of students remained at the same performance level from one year to the next, whereas with the multilevel linear growth model, students' scores revealed small, but statistically meaningful, growth year to year. The article concludes by noting advantages and disadvantages of these models to characterize growth and their implications for policy and practice.

Modeling Reading Growth in Grades 3 to 5 With an Alternate Assessment

“…Tindal and colleagues also reported large variance in intercept and growth estimates, suggesting substantial differences between students in terms of initial achievement and rates of growth. The TM, OLS, and GTS approaches evaluated in previous research, while innovative in that they modeled changes in academic performance of SWSCD across grades, appear too insensitive, whereas the MLLGM applied by Tindal et al (2016), which was tied to a common scale, demonstrated growth that was small and incremental.…”

“…Tindal et al (2016) analyzed reading growth for SWSCD in Grades 3 to 5 over the 2008–2010 school years, comparing a TM with a multi-level linear growth model (MLLGM). When applying the TM method, Tindal and colleagues found that the majority of students remained at the same performance level annually.…”

“…Although a number of studies have reviewed the academic growth for students with disabilities using various measures, relatively few have documented growth for SWSCD who participate in AA-AAS (Dunn et al, 2012;Karvonen et al, 2013;Tindal et al, 2016). Dunn et al (2012) evaluated six modeling approaches for alternate assessments using a transition matrix (TM) approach, comparing a student's performance level (e.g., above proficient, proficient, nearing proficient, well below proficient) to successive changes in annual classification.…”

“…The researchers collapsed analyses across grades to address small sample size and acknowledged their inability to model growth beyond a 2-year span due to missing data. Tindal et al (2016) analyzed reading growth for SWSCD in Grades 3 to 5 over the 2008-2010 school years, comparing a TM with a multi-level linear growth model (MLLGM). When applying the TM method, Tindal and colleagues found that the majority of students remained at the same performance level annually.…”

“…Missing data pose a serious challenge to modeling growth by introducing a potential source of systematic bias that threatens model-based inferences and generalizability (Allison, 2002; McKnight, McKnight, Sidani, & Figueredo, 2007; Schafer & Graham, 2002). Missing data are pervasive in longitudinal AA-AAS data sets, with as much as 75% attrition documented across Grades 3 to 8 (Saven et al, 2016; Tindal, Nese, Farley, Saven, & Elliott, 2016). The missingness can partially be attributed to students switching between the general and alternate assessments between years (Saven et al, 2016).…”