Online Program

Friday, February 21
CS13 Organizational Impact Fri, Feb 21, 3:15 PM - 4:45 PM
Bayshore V

Beyond Reproducibility: A Framework for an Accountable Data Analysis Process (ADAP) (302762)

*Jonathan A. Gelfond, UT Health Science Center San Antonio 
Elizabeth Heitman, Center for Biomedical Ethics and Society, Vanderbilt University  
Craig M Klugman, Department of Health Sciences, DePaul University 
Christopher Louden, UT Health Science Center San Antonio 
Bradley H Pollock, UT Health Science Center San Antonio 

Keywords: reproducibility, ethics, accountability

Data analysis is essential to the scientific process, and although recent advances in promoting reproducibility and reporting standards have made some improvements, the data analysis process remains insufficiently documented and susceptible to unintentional errors, bias, and even fraud. A shortcoming of computational reproducibility is that it fails to consider the human interactions within the research team that can affect the analytical process. Comprehensively accounting for the full process of analysis requires not only records of the data flow, but also records of communications among the research team. We propose a novel framework for capturing this analytical narrative called the accountable data analysis process (ADAP). The framework consists of data structures that characterize communications and sustain accountability, which most current analytical practices lack. We discuss the design, advantages, disadvantages, and challenges in implementing this type of system in the context of team science in such fields as clinical and translational research, but the approach is generalizable to multiple scientific disciplines.