Reproducibility and Transparency for quantitative Research

Actions Panel

Reproducibility and Transparency for quantitative Research

Reproducibility and Transparency for quantitative research: Like the layers of an onion. Professor Ben Marwick

By Data and Decision Science Network

Date and time

Thu, 28 Apr 2022 5:00 PM - 6:30 PM PDT

Location

Online

About this event

The successful reproducibility of research is fundamental to its reliability, usefulness and value to the research community and to society. However, as the computational complexity of research increases, methods and tools for ensuring the reproducibility are yet to become mainstream. In this talk I will describe an emerging consensus on ways of improving the computational reproducibility of social and natural science research. I define a research result as reproducible when the same computational analysis steps performed on the same dataset consistently produces the same answer. This is possible when authors provide all the necessary data and the computer codes for another person to run the analysis again, re-creating the same results presented by the authors. I will describe a 'layered reproducibility' approach to organise the various technologies available to enhance reproducibility in terms of effort to implement and payoff to the R-using researcher. This approach has five layers and I will present their corresponding tools and technologies. The outer layers are relatively easy to use and share, such as a single R script file. The inner layers, such as workflow management and computational environment capture, are more complex to implement for the average researcher, and have a higher payoff in improving the long-term reproducibility of the research. My observations of how researchers are tackling reproducibility indicates that for most researchers, a compendium approach to making their work reproducible is the optimum combination of effort and reward.

Organised by

Sales Ended