*3.5. Data Analysis*

The data were analyzed using SPSS version 25. The demographic properties of the sample were first described followed by a descriptive overview of the three variables SCORE, behavior\_manual, and behavior\_tracked. The proportion of participants that received perfect scores was then reported. A perfect score means that a participant identified all 11 emails correctly, or used all phishing identifiers assessed by the variables behavior\_manual and behavior\_tracked, respectively.

Next, Kruskal–Wallis H tests were used, with pairwise Mann–Whitney U test with Bonferroni correction as post hoc procedure, to identify significant between-group differences. Kruskal–Wallis H test performed on three or more samples will return a significant result if at least one sample is different from the others. In such a case, the Mann–Whitney U test with Bonferroni correction is used between all pairs in the sample to analyze what individual samples that are different from each other. Kruskal–Wallis H test was used over ANOVA because the data must show a normal distribution for ANOVA to be robust, and most samples did not in this case [46]. The conventional significance level of 0.05 is used throughout this paper.
