Hands on reproducible analysis of neuroimaging data: Nov. 2-3, UCSD

These lectures and hands-on exercises are a part of the training curriculum from the SfN 2018 training course ran by ReproNim (Reproducible Neuroimaging) Center. Selected materials are taylored for this course and cover only some sections within the full-day, but otherwise really compressed, event schedule. Please visit ReproNim: Teach for more materials, which we will also reference within specific lessons here.

Introduction

The term “reproducibility” conjures a mental image of dedicated systems conducting automated and repeatable computations. However, you can embrace reproducibility as a principle to apply to your day-to-day research activities. Neuroimaging is a heavily data and software driven field of science. As a result, by learning more tricks and techniques of the tools that you already use daily, you will discover ways to not only improve efficiency but also increase the reproducibility of your research.

To some degree, reproducibility requires knowledge of what, when, and how any particular analysis was carried out. Therefore the lessons in this module will focus on helping to answer those questions, while going from how “black box” shell could provide you valuable record of your activities, over to use of complete computational environments where versioning and origin information about each component is either exactly prescribed or just could be identified, and then to entire (simple but complete) data analysis from raw data while maintaining a complete and unambiguous provenance of all actions and access to all components of the study (code, data, computational environments).