Many higher-education efforts focus on an individual institution’s student records. The PAR Framework leverages the efficacy of federating and aggregating independent institutional datasets into a single data resource. By systematically applying a variety of exploratory, inferential, and descriptive techniques, PAR looks for patterns of risk and then mitigates those risks with the appropriate interventions at points of need.
The PAR Framework is committed to:
- Building open collaboration between 2-year, 4-year, public, private, proprietary, traditional, and progressive institutions for sharing data, analyses, and findings. Members share what they have learned about responding to challenges as a result of data, as well as successful practices that have been validated as a result of analyses.
- Developing and maintaining common data definitions; Identifying and defining common variables that apply across U.S. higher-education institutions and systems are the first steps in developing valid and reliable tools and interventions that can be utilized across the U.S. post-secondary system.
- Creating a scalable multi-institutional database that yields meaningful benchmarks for PAR institutional members and offers common, cross-institutional metrics for accountability that consider student outcomes.
- Providing affordable access to sophisticated analytical resources enabling PAR member institutions to benefit from predictive analytics, outcomes benchmarks, data warehousing, “data hygiene,” intervention measurement, and a shared talent pool.
- Advancing the efficacy of student interventions: Measures, benchmarks, and predictive models–based on common definitions–enable members to evaluate retention strategies in a variety of institutional settings and allow for measurement of intervention practices at a national level.
The PAR Framework enables analyses of 2 million de-identified student records and more than 20 million course-level records to look for patterns that can warn of risks before problems emerge. Analyses conducted on a sample size of this magnitude can offer confidence intervals in excess of 90 percent.