Project Details
Projekt
METEOR 2.0 – MastEring community knowledge-driven robusTnEss analysis in cOgnitive neuRoscience
Applicants
Professor Dr. Stefan Debener; Dr. Carsten Gießing; Professorin Dr. Andrea Hildebrandt; Professorin Dr. Christiane M. Thiel
Subject Area
Biological Psychology and Cognitive Neuroscience
Personality Psychology, Clinical and Medical Psychology, Methodology
Personality Psychology, Clinical and Medical Psychology, Methodology
Term
since 2021
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 464552782
A crucial prerequisite for the replication of brain-cognition associations obtained in or out of the laboratory with stationary and mobile neuroimaging methods is the rank order stability of neural parameters derived from noisy and complex signal recordings using different analysis methods. Given the complexity of neuroimaging data, a very large number of potential analysis pipelines are conceivable in cognitive neuroscience. Robustness is a critical requirement for replication using different data and different analysis pipelines. However, robustness analysis can only be designed with extensive knowledge of the potential data analysis options proposed in the field. In METEOR 1.0, we argued that the knowledge space of analytical choices is not traceable for individual scientists. To alleviate this problem and to promote robustness / multiverse analysis in neuroimaging (mobile EEG and graph theory based fMRI), in METEOR 1.0 we created a knowledge space of analytical choices for two neuroimaging modalities that are particularly affected by noisy and complex data structure. We have also advanced machine learning based solutions to cope with the overwhelming number of possible forks in the multiverse pipeline. In METEOR 2.0, we aim to optimize the multiverse analysis methods and tools we have developed so far by also ensuring their sustainability and testing them for plausibility, feasibility, acceptance and their positive and potentially undesirable uptake. In particular, we plan to automate the updating of the generated knowledge spaces, to extend currently available multiverse analysis methods by applying sloppy and stiff parameter analysis to deflated the multiverse, and to create a modular, multi-purpose METEOR toolbox. The toolbox will allow 1) defining the multiverse based on a comprehensive expert-rated knowledge space, 2) deflating the multiverse based on knowledge from sensitivity analysis, 3) performing multiverse analysis using machine learning when the number of pipelines is overwhelming, and 4) visualizing the variability of results and/or integrating results across the multiverse of pipelines. We will also work to standardize and disseminate study reporting practices. This will contribute to more robust and replicable cognitive neuroscience.
DFG Programme
Priority Programmes
