Reducing the efforts to create reproducible analysis code with FieldTrip

Abstract

The analysis of EEG and MEG data typically requires a lengthy and complicated sequence of analysis steps, often requiring large amounts of computations, which are ideally represented in analysis scripts. These scripts are often written by researchers without formal training in computer science, resulting in the quality and readability of these analysis scripts to be highly dependent on individual coding expertise and style. Even though the computational outcomes and interpretation of the results can be correct, the inconsistent style and quality of analysis scripts make reviewing the details of the analysis difficult for other researchers that are either involved in the study or not, and the quality of the scripts might compromise the reproducibility of obtained results. This paper describes the design and implementation of a strategy that allows complete reproduction of MATLAB-based scripts with little extra efforts on behalf of the user, which we have implemented as part of the FieldTrip toolbox. Starting from the researchers' idiosyncratic pipeline scripts, this new functionality allows researchers to automatically create and publish analysis pipeline scripts in a standardized format, along with all relevant intermediate data. We demonstrate the functionality and validate its effectiveness by applying it to the analysis of a recently published MEG study.

Good Scientific Practice in MEEG Research: Progress and Perspectives

“…With some discipline, these tools allow keeping track of every change made to the code, and allows efficient collaboration at scale. Furthermore, more and more analysis toolboxes such as those presented below offer script generators to improve reproducibility and generate code that is easy to understand and adapt (Es et al, 2021), which reduces the burden on the researcher.…”