Does the Procedure Matter? Applying a Multiverse Analysis Approach to Negative Emotion Differentiation

Negative emotion differentiation (NED) is frequently assessed using momentary emotion ratings in ambulatory assessment studies. However, researchers differ in how they preprocess data in studies on NED, and this variation may affect empirical findings. The present research scrutinized whether decisions in NED data processing affect the robustness of NED’s relationships with adaptive outcomes using multiverse analysis—that is, conducting the analysis of interest across all reasonable combinations of methodological decisions. We included decisions on the compliance threshold for exclusion, the inclusion of occasions, and how the NED index was calculated. The analyses of interest were the bivariate between-person correlation between NED and depression and, using multilevel analysis, the buffering effect of NED on daily stress reactivity. For both analyses of interest, separate multiverse analyses were conducted on three ambulatory assessment data sets (163 to 406 participants, 11,876 to 21,552 occasions) collected in Germany between 2020 and 2024. The results indicated that the bivariate between-person correlation between NED and depression was not robust in any of the data sets. The buffering effect of NED on daily stress reactivity was robust in one data set. The impact of specific decisions varied across data sets and analyses of interest. With the exception of one decision, each decision was identified as the most influential in at least one multiverse analysis. However, there was no systematic overall pattern in how the decisions affected the outcomes. The results of the multiverse analyses highlight the importance of research transparency and careful data preparation, as methodological decisions may affect empirical findings on NED. Whether the results generalize to other analyses of interest and to more diverse or non-German samples remains to be determined.