More than one way to analyse a dataset

The way that different research teams approach the analysis of raw neuroimaging data varies and is becoming increasingly complex. To estimate the potential variability caused by different analysis workflows in the results of functional magnetic resonance imaging, a single dataset was distributed to 70 teams around the globe, with each laboratory testing 9 ex-ante hypotheses. Every group approached the task differently, leading to 70 different workflows being established in the course of the analysis. This flexibility in workflow resulted in sizeable differences in the hypothesis test results. The study emphasizes the importance for research teams of validating and sharing their analysis workflows when exploring such datasets.