62 Chapter 3 within-person level, justifying the multilevel design of the current study and supporting a multilevel analysis approach. In addition, we tested whether the slope between the independent variable (proactive vitality management) and the dependent variable (i.e., creative work performance) varied across respondents. The slope variance was significant (variance component = .13, p < .001), justifying the introduction of person-level variables (i.e., selfinsight and social support for creativity) into the analyses to test cross-level interaction effects. Testing (cross-level) moderation hypotheses requires the inclusion of the main effects of the moderators into the analysis (Aiken, West, & Reno, 1991). Therefore, we controlled for the main effects of self-insight and social support for creativity when testing hypothesis 2 and 3. We conducted multilevel regression analysis based on four nested models introducing successively the intercept (Null model), the predictor (Model 1), the two moderators (Model 2), and their two hypothesized cross-level interaction effects (Model 3). In the multilevel regression analyses, the person-level variables (i.e., self-insight, social support for creativity, proactive personality, and creative requirement) were grandmean centered, and the week-level predictor (proactive vitality management) and control variable ( job autonomy) were group-mean centered (Ohly, Sonnentag, Niessen, & Zapf, 2010). Both the outcome variable (i.e., weekly creative work performance) and the control variable ‘time’ remained uncentered. However, as the time variable proved to be an insignificant predictor of creative work performance and explained no variance, we decided to exclude this variable from further analyses (Hox, 2010). RESULTS Descriptive Statistics Means, standard deviations and correlations of the variables in the current study are presented in Table 1.
RkJQdWJsaXNoZXIy MTk4NDMw