45 Conceptualization and Measurement of Proactive Vitality Management detachment, both on general and day-level (r = .03, p = .473 and r = .10, p = .111, respectively), indicating support for hypothesis 3b (Table 2 and 3). In the general-level sample, we found positive relationships between PVM and cognitive liveliness (r = .48, p < .001) and work engagement (r = .62, p < .001). Moreover, PVM was negatively related to exhaustion (r = -.43, p < .001). In addition, PVMwas positively related to vigor (r = .64, p < .001), and negatively related to fatigue (r = -.48, p < .001) in the daylevel study. Hypothesis 4a, 4b and 4c were thus supported (Table 2 and 3). However, due to the relatively high correlations between PVM and work engagement (general-level) and vigor (day-level), we conducted additional analyses to further test whether the constructs could be empirically discriminated. First, the Average Variance Extracted (AVE) estimates of PVM and work engagement (.51 and .64) indicate that the two constructs can indeed be distinguished, as both estimates were greater than the shared variance (i.e., squared correlation) between the two factors (.38) (Farrell, 2010; Fornell & Larcker, 1981). Similar results were found in the day-level study, in which the AVE estimates of PVM and vigor (.56 and .71) were both greater than their shared variance estimate (.41). Second, we conducted CFAs to compare a model in which the items of each construct load on their own respective latent factor, versus a model in which all items load on one overall latent factor. In the general-level data, the model in which the indicators of work engagement and PVM loaded on two separate factors fit the data significantly better than the one-factor model (Δχ² = 749.42, Δdf = 1, p < .001). Moreover, in the day-level data, themodel inwhich the indicators of vigor and PVM loaded on two separate factors also fit the data considerably better than the one-factormodel (Δχ² = 284.37, Δdf = 1, p < .001). Taken together, these results clearly show that PVMcan be empirically distinguished fromvigor andwork engagement. Finally, the results provided support for criterion validity, as we found positive relationships between PVM and in-role work performance (r = .30, p < .001) as well as creative work performance (r = .37, p < .001) on the general-level. Similar results were found in the day-level sample (r = .50, p < .001 and r = .40, p < .001, respectively). Furthermore, we found a significant, positive relationship between PVM and cognitive performance (scores on the RAT) in the day-level study (r = .14, p < .05). This means that hypothesis 5a, 5b and 5c were supported as well (Table 2 and 3). Overall, the findings of Study 3 show that the use of PVM relates to relevant variables in its nomological network, and provide support for convergent, discriminant, and criterion validity. 2
RkJQdWJsaXNoZXIy MTk4NDMw