115 Creating A Creative State of Mind Work engagement. We used the adapted version (Breevaart et al., 2012) of the nine-item Utrecht Work Engagement Scale (UWES; Schaufeli et al., 2006) to measure daily work engagement. Example items are “Today at work, I felt bursting with energy” (vigor), “Today, I was inspired by my job” (dedication) and “Today, I was immersed in my work” (absorption) (1 = totally disagree, 7 = totally agree). The average Cronbach’s alpha over the five days was .96. Strategy of Analysis In our data, daily measurements were nested within persons. Therefore, we tested our hypotheses using multilevel analysis (HLM 7.01 software; Raudenbush et al., 2013). For each variable, we calculated the intra-class correlation (ICC) in order to obtain the percentage of variance that can be attributed to the within-person level. The resulting percentages (i.e., 65% for originality, 41% for fluency, 44% for mindfulness, and 31% for proactive vitality management) justified the use of a multilevel design. The outcome variables (i.e., fluency and originality) and the control variable ‘time’ remained uncentered, while proactive vitality management and mindfulness were centered at each individual’s mean value (Ohly et al., 2010). To take into account the potential carry-over effects of one’s prior level of mindfulness and creativity, we created lagged variables and performed amore stringent test by including these previous-daymeasures of the mediator (i.e., mindfulness) and outcomes (i.e., fluency and originality) in our analyses. In the multilevel analyses with either fluency or originality as the outcome, we first entered the control variables time and work engagement and the lagged variable of the respective outcome (Model 1). In the next step, we entered mindfulness (Model 2). In addition, we tested the indirect effect of proactive vitality management on creative performance through mindfulness using the Monte Carlo method for assessing multilevel mediation (Preacher & Selig, 2010). To ensure robustness of our findings, we followed the recommendation to also test our hypotheses without including any control variables (Becker et al., 2016). Without the control variables, the results supported our hypotheses in the same way – there were no differences in the direction or significance of the resulting relationships.3 3 The results without the control variables may be requested from the first author. 5
RkJQdWJsaXNoZXIy MTk4NDMw