Newly developed computational interface helps data novices interact with in-depth data
August 13, 2025
ANN ARBOR — A team at the University of Michigan’s Institute for Social Research has developed a new computational interface that could allow for more people to apply multilevel regression and poststratification to their data, even if they don’t have a statistics background or experience with R programming language
Yajuan Si, a research associate professor in ISR’s Survey Research Center (SRC), and Toan Tran, a research software engineer, led the effort, which was launched as a way for analysts to apply statistical methods to data such as COVID-19 test records, which Si considers one of the ideal use cases.
“Patients who have elective surgery appointments with the hospital provide COVID test samples, and hospitals can then analyze these data with this interface to generate community-wide findings,” she said. “The data paired with the interface can help predict the clinical burden of public health emergencies.”
Si and Toan refer to the interface, available online as an app and as an installable R package, as “shinymrp.” The tool should open the door for non-experts who still need to work with crucial data to make important decisions, such as those that would need to be made in a timely fashion during an infectious disease outbreak to manage care plans and resources across an entire hospital.
“Hospital test records are part of medical records, and you have to follow HIPAA privacy regulations, which prevents researchers like us from easily accessing that data,” said Si. “But for those who work inside the hospital, they have access to the data. We want them to use our tools and run their data analysis locally. In that sense, we don’t need to see their data, but they can run our approach.”
Si explored the hospital use case in some depth in a paper to be published later this year titled “Multilevel Regression and Poststratification Interface: An Application to Track Community-level COVID-19 Viral Transmission.” In it, she demonstrates the effectiveness of the app and shows how COVID data can be explored using the interface. Beyond infectious disease monitoring, the MRP interface can be applied to broad data analyses in health and social science research. It accommodates time-varying and cross-sectional data, continuous and binary outcomes, and supports subgroup analyses across demographic and geographic domains. Users can customly specify models and input data sources.
After two years in development the app is fully usable now, but development still continues. As a programmer, Toan has his eyes on numerous new features that would enhance the app’s usability.
“For example, we can fit more sophisticated models by looking at spatial correlation among different geographic areas,” he said. “So those who are closer to each other geographically, many have similar patterns with COVID infection. It is a refining process of different modeling techniques.”
The app, developed with funds from a research grant, is available here. Si and Toan have developed both a detailed user guide and a video walkthrough to help any potential users navigate the tool.
Contact: Jon Meerdink ([email protected])