Using hierarchical modeling to determine grade proficiency on the New England Common Assessment Program testing
The purpose of this study is to find the factors that are important for student performance on the New England Common Assessment Program (NECAP). More specifically, the school, socio-economic, and teaching factors will be studied to determine what had an impact on the rate of mathematics proficiency for students in each grade throughout the state of Rhode Island. The NECAP is the assessment program that the Rhode Island Department of Education (RIDE) designed to meet the specifications of the No Child Left Behind Act (NCLB) enacted by President Bush in 2002. Grade level information was collected from the RIDE's Frequently Requested Educational Data (FRED), which has been displayed on RIDE's website for the public to access. Information such as race, socio-economic level, and gender was collected and used to understand the demographic importance in regards to reaching proficient on the NECAP. In addition, school level information such as attendance rates, mobility rates, and faculty composition also were collected to determine their influence on proficiency. These variables were collected for four school years—2006–07, 2007–08, 2008–09 and 2009–10—for every public and charter school in the state of Rhode Island for the 3rd, 4 th, 5th, 6th, 7th, and 8th grades. In modeling the rate of proficiency, or the percentage of students who have reached a level of proficient, on the mathematics assessment, the researcher investigated the following models to determine the best possible non-hierarchical model: the Poisson Regression, the Poisson Regression for Rates, the Poisson Regression for Rates Adjusting for Dispersion, and the Negative Binomial Regression for Rates. The selected best non-hierarchical model was chosen to be further modeled as a hierarchical model. Once this was completed, the hierarchical and non-hierarchical models were compared using goodness of fit assessments. In conclusion, the Poisson Regression for Rates Adjusting for Dispersion was the best possible non-hierarchical model. When compared to the hierarchical version, the Hierarchical Poisson Regression for Rates Adjusting for Dispersion was the best possible model to predict the rate of proficiency on the mathematics section of the NECAP testing.
Educational tests & measurements|Statistics
Kristen E Roland,
"Using hierarchical modeling to determine grade proficiency on the New England Common Assessment Program testing"
Dissertations and Master's Theses (Campus Access).