Search for: All records

Multiverse analyses involve conducting all combinations of reasonable choices in a data analysis process. A reader of a study containing a multiverse analysis might question—are all the choices included in the multiverse reasonable and equally justifiable? How much do results vary if we make different choices in the analysis process? In this work, we identify principles for validating the composition of, and interpreting the uncertainty in, the results of a multiverse analysis. We present Milliways, a novel interactive visualisation system to support principled evaluation of multiverse analyses. Milliways provides interlinked panels presenting result distributions, individual analysis composition, multiverse code specification, and data summaries. Milliways supports interactions to sort, filter and aggregate results based on the analysis specification to identify decisions in the analysis process to which the results are sensitive. To represent the two qualitatively different types of uncertainty that arise in multiverse analyses—probabilistic uncertainty from estimating unknown quantities of interest such as regression coefficients, and possibilistic uncertainty from choices in the data analysis—Milliways uses consonance curves and probability boxes. Through an evaluative study with five users familiar with multiverse analysis, we demonstrate how Milliways can support multiverse analysis tasks, including a principled assessment of the results of a multiverse analysis. 

Recent studies have shown that users of visual analytics tools can have difficulty distinguishing robust findings in the data from statistical noise, but the true extent of this problem is likely dependent on both the incentive structure motivating their decisions, and the ways that uncertainty and variability are (or are not) represented in visualisations. In this work, we perform a crowd-sourced study measuring decision-making quality in visual analytics, testing both an explicit structure of incentives designed to reward cautious decision-making as well as a variety of designs for communicating uncertainty. We find that, while participants are unable to perfectly control for false discoveries as well as idealised statistical models such as the Benjamini-Hochberg, certain forms of uncertainty visualisations can improve the quality of participants’ decisions and lead to fewer false discoveries than not correcting for multiple comparisons. We conclude with a call for researchers to further explore visual analytics decision quality under different decision-making contexts, and for designers to directly present uncertainty and reliability information to users of visual analytics tools. The supplementary materials are available at: https://osf.io/xtsfz/.