Using Information Principles for Guiding Data Analysis

Presented by: Śaunak Sen, PhD, Professor and Chief, Division of Biostatistics

Location: Freeman Auditorium, 3rd floor of the 930 Madison Building, 11/6/25 12noon – 1pm CT

Remote/Zoom: https://tennesseehipaa.zoom.us/webinar/register/WN_5ZBwl0ygTZy9OdG2VFMlwQ

Food: Pizza will be available starting at 11:45AM on a first-come, first-serve basis.

Join us this Fall for the Statistical Reasoning in Biomedical Research Seminar Series by the Division of Biostatistics in the Department of Preventive Medicine!  The third talk is titled “Using Information Principles for Guiding Data Analysis". Individual data points often naturally belong to groups.  For example, voters belong to precincts, laboratory mice may be classified into strains, and inpatient records may be grouped by the admitting hospital.  In many settings the question is: should we analyze the individual-level data (if available) or the aggregated data at the group level?  Using a practical example, we will use the following information principles to think about the question:

• Information in processed data cannot exceed that in the original data.

• The information in the sample mean is proportional to sample size and inversely proportion to the variance.

• If we have three or more groups, sharing information between groups usually improves estimates.

These general principles are broadly useful for guiding practical data analysis.

Mark your calendar: Additional talks in this series are scheduled for 11/13/25, and 12/04/25. (all 12noon – 1pm CT at the Freeman Auditorium)