*2.3. Data Analysis*

To assess normality, distributions of all analyzed variables were visually assessed and tested using Kolmogorov–Smirnov tests. Indeed, this approach revealed that all of the tested variables significantly deviated from the normal distribution in both sample groups (all *p*s < 0.007). Accordingly, predominantly non-parametric as well as robust approaches were applied throughout the entire analysis. In order to extract a meaningful scale to express a rise in increased and more unhealthy food intake during the COVID-19 pandemic, a factorial analysis was applied to the 10 items measuring COVID-19-specific eating behavior (DCSEB). Self-generated items for *dysfunctional safety behavior* have been intensively discussed in previous studies by our group (please see [33]). Cronbach's α for *dysfunctional safety behavior* in the current sample was 0.794.

To test univariate associations between COVID-19-related variables—*generalized anxiety*, *depression*, *dysfunctional COVID-19-related eating behavior*, and *dysfunctional safety behavior*—Spearman correlation coefficients were computed. To further explore whether obesity surgery had an influence on the respective psychopathological dimension (PHQ-8, GAD-7), COVID-19-related fear, and dysfunctional COVID-19-related eating behavior (DCSEB), group differences (with vs. without surgery) were assessed via Mann–Whitney U tests. Separate robust regression analyses—as implemented in the R package *robustbase* [34]—were then computed to assess whether the associations between *DCSEB* and COVID-19-related fear, depression, and anxiety symptoms (PHQ-8 and GAD-7) are moderated by obesity surgery. To do so, the respective psychological variable, the group variable (*with* and *without* obesity surgery), as well as their interaction coefficients were regressed on *DCSEB*. A full summary of regression coefficients is provided in the Supplemental Materials. The data were analyzed using IBM Statistics SPSS 26 (New York, NY, USA) and R (3.6.3).
