Newsroom

The Science of Sleep: A New Method for Analyzing Experimental Time Series Data

August 1st, 2017

Posted in:

Scott Bruce, a PhD student at the Fox School of Business, publishes in top statistics research journal, Biometrics.

Researchers are up to their elbows in data. But new methods are needed to address the increasing size and complexity of modern data structures, especially those with time-varying dynamics.

“Frequently, with biomedical experiments, you have time series data from many different participants, along with other clinical and behavioral information for each,” explains Scott Bruce, a fifth-year PhD student in the Department of Statistical Science at the Fox School.

“However,” he continues, “much of the statistical literature focuses on the analysis of a single time series, so there is a need for new theory and methods that can analyze more complex data generated from these kinds of modern biomedical experiments.”

Bruce’s article on his recent research—under the advisement of Dr. Robert Krafty (University of Pittsburgh) and Dr. Cheng Yong Tang (Temple University)—was accepted for publication by the prestigious statistics journal, Biometrics (and appears online through “Biometrics Early View”). He has developed a new method called “conditional adaptive Bayesian spectrum analysis,” or CABS, which can be used to analyze associations between the dynamics of time series data and other data of interest.

The University of Pittsburgh’s AgeWise Caregiver Study, which examined the relationship between stress and sleep in older adults serving as the primary caregiver for an ill spouse, was the motivation for Bruce’s research. Study participants were monitored during a night of sleep to obtain their heart rate variability time series data, and they completed a questionnaire in order to formulate an individualized sleep score (the Pittsburg Sleep Quality Index).

The goal was to measure the association between physiological stress captured in the heart rate variability data and sleep quality measured by the PSQI score.

“This new method allows you to analyze these associations between the covariates and the time series,” explains Bruce. “It properly reflects the temporally-evolving nature of the relationship and helps us better understand dynamic biological processes. Researchers and practitioners who study time series data in conjunction with other types of data will be interested in this work.”

Bruce and his collaborators also developed user-friendly functions available in MATLAB that allow researchers and practitioners to use CABS in their own work. It’s available at the Biometrics website.

Learn more about Fox School Research.
For more stories and news, follow the Fox School on Twitter, Facebook, and Instagram.

See All Idea Marketplace Posts