|
Using Hierarchical Linear Growth Models to Investigate Reading Achievement Gaps
|
| Presenter(s):
|
| Tammiee Dickenson, University of South Carolina, tsdicken@mailbox.sc.edu
|
| Abstract:
This study illustrates the use of hierarchical linear modeling to investigate achievement growth, including examination of demographic achievement gaps. The sample consisted of 1615 students in 46 schools that participated in the South Carolina Reading First (SCRF) Initiative during the first three years. The Stanford Reading First assessment was administered to SCRF students in grades 1-3 in the fall and spring of each school year. A three-level hierarchical linear model was used to model growth in achievement for students who participated in all three years. A quadratic term was included to account for change in growth rate over time. Comparisons were made for various subgroups of the population in terms of both initial achievement and growth rates. Achievement gap changes were analyzed by demographic subgroups and by whether students received intervention services to investigate differential impacts. This study has relevance to evaluators of multi-year programs intended to impact student achievement.
|
|
Using State Archival Data to Estimate School-level Causal Effects: Lessons Learned From a Multi-state Quasi-experimentaL Study
|
| Presenter(s):
|
| Aikaterini Passa, ICF International, apassa@icfi.com
|
| Allan Porowski, ICF International, aporowski@icfi.com
|
| Susan Siegel, Communities In Schools, siegels@cisnet.org
|
| Abstract:
As with experimental studies, the quality, validity, and reliability of quasi-experimental studies depend in large part on the rigor with which data is collected, outcome measures are selected, possible confounds and biases are identified and addressed, and appropriate statistical techniques are applied. This presentation will cover the steps taken and the criteria established in conducting the Communities In Schools (CIS) school-level quasi-experimental study across seven states: Texas, Florida, Georgia, Pennsylvania, Michigan, Washington, and North Carolina. Suggestions for researchers will be presented, including how to conduct a high quality school-level quasi-experimental study using state archival data. School-level data were collected from each State's Department of Education website, and propensity score matching was used to match CIS sites with comparison schools. Ultimately, propensity score matching produced well-matched comparison schools, demonstrating the application of this technique in the conduct of rigorous, retrospective evaluations.
|
| |