Workshops & Training Offered to Statisticians Across Campus to Become Versed in Both Frequentist and Bayesian Methods

Creating innovative initiatives and facilitating interdisciplinary collaborations is a hallmark of the U-M research enterprise. Data analysts and statisticians who find themselves without support from other statisticians now have the Applied Biostatistical Sciences Network (ABS), established by Michigan Institute for Clinical & Health Research (MICHR), to bring together biostatisticians and research investigators to collaborate, share ideas and learn cutting-edge research methods.

From statisticians and epidemiologists to data scientists and researchers across campus, ABS Network membership has grown to more than 260 members in its first two years.

Training and workshops have included well-received offerings including: Biomarkers-Driven Clinical Trial Design; Statisticians Are from Mars, Clinicians Are from Venus; and the Bayesian Methods.

“The ABS network mission is to improve health and health research by fostering partnerships among statisticians, data analysts, epidemiologists, data scientists and researchers across the university,” said Shokoufeh Khalatbari, MICHR’s Biostatistics Program Manager. “One shining achievement is our Bayesian methods workshop series for statisticians across campus.”

There are two schools of statistical inference – Frequentist and Bayesian – which differ in how probability is used.

In the Frequentist approach, statisticians use only the data to draw conclusions.
In the Bayesian approach, statisticians use the data, as well as prior knowledge about the event or conditions related to the event, to make decisions.

While the differences may seem subtle, in practice they are vastly different statistical approaches for analyzing data. The majority of biostatisticians, epidemiologists, and researchers at U-M have expertise only in Frequentist approaches, primarily because formalized training and mentorship opportunities in Bayesian methodology were severely lacking.

To address this gap, the ABS Network established the Bayesian workshop series. ABS partnered with instructors from the School of Public Health to design the workshops, with the aim of increasing the number of biostatisticians across U-M who could apply Bayesian methods. This training series was tailored to fit into the schedule of a full-time staff biostatistician member and included post-workshop mentorship by the instructors.

Unlike traditional trainings that are commonly lecture-based, the training aimed for a unique, immersive experience where statisticians tackled defined projects of their choosing throughout the entire training. In this way, participants gained a deep understanding of the topic and are now able to apply the methodology. Following each training session, attendees were surveyed to gather information on their confidence level in performing several tasks. Data collected by MICHR’s Biostatistics team showed increasing levels of confidence as the classes and topics progressed.

The participants were required to present a capstone project to complete the training, supported by two capstone support clinics. Twenty people successfully completed the workshops. Participants presented their capstone projects and graduated in February 2020 (see attached picture). Many of the capstones are now being developed into manuscripts that will be submitted to peer-reviewed journals. There are also plans to collaborate with the MICHR Education & Mentoring group to develop a training module and associated resources to be shared with other Clinical & Translational Science Award (CTSA) hubs across the country. Results of this effort will be disseminated at conferences.

Learn more about the ABS Network.