A Survey of Tasks and Visualizations in Multiverse Analysis Reports

Abstract

Analysing data from experiments is a complex, multi‐step process, often with multiple defensible choices available at each step. While analysts often report a single analysis without documenting how it was chosen, this can cause serious transparency and methodological issues. To make the sensitivity of analysis results to analytical choices transparent, some statisticians and methodologists advocate the use of ‘multiverse analysis’: reporting the full range of outcomes that result from all combinations of defensible analytic choices. Summarizing this combinatorial explosion of statistical results presents unique challenges; several approaches to visualizing the output of multiverse analyses have been proposed across a variety of fields (e.g. psychology, statistics, economics, neuroscience). In this article, we (1) introduce a consistent conceptual framework and terminology for multiverse analyses that can be applied across fields; (2) identify the tasks researchers try to accomplish when visualizing multiverse analyses and (3) classify multiverse visualizations into ‘archetypes’, assessing how well each archetype supports each task. Our work sets a foundation for subsequent research on developing visualization tools and techniques to support multiverse analysis and its reporting.

Open and reproducible neuroimaging: from study inception to publication

“…The converging findings of these studies across modalities suggest that it is important to test the robustness of reported results to specific analytical choices. One proposed solution to tackle the analytical variability, where many different analytical approaches are compared, is multiverse analysis (Hall et al 2022). There are two broad types of multiverse tools.…”

Analysis reproducibility in mental health research: challenges and solutions

“…One solution that has been suggested is "multiverse analysis" (99,100), in which a range of theoretically and statistically reasonable analysis pipelines are tested and reported. This concept has been used across fields, sometimes under different names (e.g., specification curve analysis (101,102)).…”

“…Third, multiverse analysis is complex, making it challenging to integrate, visualize, and summarize findings and conclusions. Diverse approaches to visualization and reporting have been developed (99,101,107,108). One statistical approach for interpretation of multiverse analysis in fMRI is a "consensus analysis", which is a type of meta-analysis over the statistical maps resulting from the different pipelines, adapted to account for their dependency (since they are based on the same data) (86,109).…”

The multiverse of data preprocessing and analysis in graph-based fMRI: A systematic literature review of analytical choices fed into a decision support tool for informed analysis

“…A full multiverse analysis should then be followed by a full report of the results from all possible combinations of forking paths. This report might be, for example a p -value distribution, a specification curve analysis and plot, or a vibration of effects plot (for review, see Hall et al, 2022). A joint conclusion can then be drawn by integrating the results across the multiverse of options (Simonsohn et al, 2020) or by reporting a multiverse variability index that quantifies the robustness of the results across the forking paths (Olsson-collentine et al, 2023).…”