Description: States, districts, and schools are increasingly using early warning indicator systems in which readily available data on student attendance, academics, and behavior are used to identify students at risk for not graduating from high school. Identified students can then be matched with interventions in an attempt to help them get on track for graduation. REL Midwest is working with three Ohio districts to develop a set of locally tailored early warning indicators for each district. Study results will provide information about the strongest indicators of four-year graduation for the students in each district, as well as the indicator thresholds at which students are most accurately identified as graduates or nongraduates.
Research Questions: For this project, REL Midwest will address the following four research questions:
What are the strongest eighth- and ninth-grade raw-data predictors of four-year high school graduation in each district?
For each of the strongest eighth- and ninth-grade raw-data predictors in each district, what is the most sensitive and specific binary indicator that can be constructed?
In each district, which of the constructed indicators developed under Research Question 2 best predict four-year graduation, individually and in combination (both within and across grade levels)?
In each district, which of the constructed indicators are most strongly associated (individually and in combination) with four-year graduation?
In each district, how accurately do the indicators and combinations of indicators identified under Research Question 3a classify graduates and nongraduates?
To what degree is there variation across districts in the strongest locally specific indicators?
Study Design: Each district has provided data on student attendance, academics, and behavior for Grades 8–12 for two cohorts of students. Regression analyses will identify the strongest eighth- and ninth-grade predictors of four-year graduation. These predictors will be converted to binary indicators using receiver operating characteristic curve analysis. A second set of regression analyses will identify the strongest binary indicators of graduation. Descriptive analysis will be used to examine the accuracy with which single and multiple indicators identify graduates and nongraduates. Results will be presented to highlight similarities and differences among the three districts.
Projected Release Date: Fall 2014
Research Alliance: Dropout Prevention Research Alliance, REL Midwest
Keywords: Dropout Prevention, Data Use and Systems, High School, Early Warning Indicators, Correlational, Descriptive Statistics, Ohio
Study-Related Products: Making Connections report.