The second hypothesis stated that high sleep quality may buffer the effects that momentary job
demands have on momentary work engagement and low sleep quality may enhance the negative effects of momentary job demands on momentary work engagement. In the analyses, the four items of work engagement were compared to the five variables (self-perceived sleep quality, self-perceived recovery, observed total sleep time, observed sleep latency and observed sleep efficiency) of sleep quality. The moderating variables are discussed separately in this section.
Self-perceived sleep quality
The results of the analysis showed that self-perceived sleep quality of the previous night directly led to an increase in DED2 (g = .22, se = .06, p = .001) and VIG2 (g = .17, se = .05, p = .001). Results of the analysis showed no direct effect between self-perceived sleep quality, DED1 (g = .06, se = .26, p =.797) and VIG1 (g = .10, se = .05, p = .051). Self-perceived sleep quality did not affect the relation between momentary job demands and the items of dedication was found (DED1: g = -.02, se = .04, p = .564; DED2:
g = -.05, se = .04, p = .184). The moderation of self-perceived sleep quality also did not affect the relation between momentary job demands and the items of vigor (VIG1: g = -.05, se = .03, p = .109; VIG2: g = -.04, se = .03, p = .155). Adding self-perceived sleep quality to the model that only included the intercept and momentary job demands improved the model fit for DED2 2*log(lh) = 122.31, p < .001) and VIG2 (D-2*log(lh) = 98.77, p < .001).
Randomization of slope was performed for momentary job demands to find out if there was a within-person day-level effect of momentary job demands on DED2. Self-perceived sleep quality was found to moderate the relation between momentary job demands and DED2 (g = -.08, se = .04, p = .050). The slope variance also revealed that within days, momentary job demands led to more DED2 (s2 = .03, se = .01 p = .006). When performing randomization of slope for momentary job demands, the model showed an improved fit (D-2*log(lh) = 3.71*) over the model without the interaction, but with the
randomization. The results of this randomization can be found in Table 5.
Finally, in order to interpret the interaction effect and direction of self-perceived sleep quality, a simple slopes test was performed. The results show that under the condition of ‘low’ self-perceived sleep quality of the previous night (one standard deviation below the sample mean), momentary job demands led to an increase in DED2 (t = 4.49, p < .001). In addition, results show that for ‘high’ sleep quality (one deviation above the sample mean), job demands did not relate to DED2 (t = 1.67, p = .118). Figure 2 shows that for ‘low’ sleep quality, DED2 increases as momentary job demands also increase. The results of the analysis performed for this hypothesis can be found in Appendix 7.2. Despite the fact that there was a positive moderation for the second item of dedication, this was in the wrong direction and only for one out of four items of work engagement. Therefore, for the moderator self-perceived sleep quality, hypothesis 2 was rejected for DED1, DED2, VIG1 and VIG2.
Table 5: Randomization of slopes in analysis of moderation of sleep quality of relation between job demands and DED2.
Base model Randomized model Interaction model
Estimate SE Z value Estimate SE Z value Estimate SE Z value Constant 5.69 .09 61.01** 5.68 .10 58.87** 5.68 .10 58.52**
Job demands .10 .02 4.86** .11 .02 4.46** .11 .02 4.51**
Sleep quality .22 .06 3.55** .18 .06 2.98* .22 .06 3.42**
Job demands x sleep
quality -.08 .04 -1.97*
-2*log(lh) 4312.98 4280.64 4276.93
D -2*log(lh) 122.32** 32.34** 3.71*
Df 1 1 2
Randomized slope
job demands .03 .01 .03 .01
Level 3 intercept
variance (person) .36 .09 .38 .09 .40 .09
Level 2 intercept
variance (day) .05 .04 .08 .04 .05 .02
Level 1 intercept
variance (moment) 1.60 .07 1.48 .07 .93 .04
Note: * p < .05, **p £ .001.
Figure 2: Simple slope of cross-level interaction between job demands, self-perceived sleep quality and DED2.
Self-perceived recovery
In Table 6 the analysis of the effects of self-perceived recovery on the relation between momentary job demands and the items of work engagement is shown. Although self-perceived recovery was related to the items of vigor (VIG1: g = .12, se = .06, p = .048; VIG2: g = .24, se = .06, p = .001), self-perceived
recovery was not found to moderate the relation between momentary job demands and any of the items of vigor (VIG1: g = .01, se = .04, p = .807; VIG2: g = -.02, se =.03, p = .478). For the second item of
dedication however, the analysis showed that more self-perceived recovery led to more DED2 (g = .30, se
= .07, p = .001). Furthermore, self-perceived was found to moderate the relationship between
momentary job demands and DED2 (g = -.09, se = .05, p = .038), in such a way that more momentary job demands would lead to less DED2 when moderated by self-perceived recovery. When self-perceived recovery was added to the model including only the intercept and momentary job demands, this improved the model fit for DED2 (D-2*log(lh) = 98.98, p < .001).
Randomization of slope was performed for this analysis to see if there was a within-person, day-level effect of momentary job demands on DED2. The randomization of slope analysis showed that for DED2, self-perceived recovery interacted with job demands (g = -.11, se = .05 p = .025), and the slope variance of the randomization showed that within days, more self-perceived recovery led to more dedication (DED2) (s2 = .03, se = .01 p = .006). Moreover, the model fit was improved (D-2*log(lh) = 4.91, p = .027) in comparison to the model which only included the randomization.
In order to find the direction of the effect of self-perceived recovery on the relation between momentary job demands and momentary work engagement, a simple slopes analysis was performed. The results of this analysis show that for ‘low’ self-perceived recovery (one standard deviation below the sample mean), more momentary job demands led to more DED2 (t = 4.80, p < .001). This result indicates that when a participant experienced ‘low’ self-perceived recovery from the previous night, momentary dedication (DED2) increased as momentary job demands also increased. In addition, results show that for ‘high’ self-perceived recovery (one standard deviation above the sample mean), no relation between job demands and DED2 was found (t = 1.55, p =.122). Figure 3 shows that as momentary job demands increase, DED2 also increases for low self-perceived recovery, indicating that the variables are related in the opposite way for DED2 then what the hypothesis states. The results of the analysis performed for this hypothesis can be found in Appendix 7.2. Thus, the tests indicated that more momentary job demands lead to more dedication (DED2) under the condition of low self-perceived recovery. Thus, hypothesis 2 with self-perceived recovery as moderator was rejected for DED1, DED2, VIG1 and VIG2.
Observed total sleep time
The analysis showed that as the total sleep time of a participant decreased, this also led to less dedication (DED2) (g
= .00, se = .00, p = .001). Observed total sleep time was unrelated to DED1, VIG1 and VIG2. However, total sleep time was not found to moderate the relationship between momentary job demands and DED2 (g = .00, se = .00, p = .349). When observed total sleep time was added to the model including only the intercept and momentary job demands, this improved the model fit for DED2 (D-2*log(lh) = 758.10, p < .001). Randomization of slope was performed for momentary job demands to find out if there was a within-person day-level effect of momentary job demands, but no day-level variance was detected.
Afterwards, a simple slopes test was performed to investigate the effect of observed total sleep time on the relation between momentary job demands and DED2. This analysis revealed that for both high (t = 7.33, p < .001) and low (t = 2.35, p = .019) levels of observed total sleep time, DED2 increased as momentary job demands also increased, but showed no real difference between the two lines. The results of the analysis performed for this hypothesis can be found in Appendix 7.2. All in all, the results showed that hypothesis 2 with observed total sleep time as moderator was only supported for DED2. It was rejected for all other items of work engagement (DED1, VIG1 and VIG2).
Figure 3: Simple slope test of cross-level interaction between momentary job demands, self-perceived recovery and DED2.
Table 6: Multilevel analyses of the daily self-perceived recovery on the relation between momentary job demands
-2*log(lh) 4285.98 4336.32 3693.25 3612.64
D -2*log(lh) 73.02** 98.98** 66.95** 84.92**
Note: In Model 2: Recovery, the D -2*log(lh) of model 2 was based on the -2*log(lh) of model 1, which can be found in Table 4. *p < .05 **p <
.001.
Observed sleep latency
The analysis showed no relation between observed sleep latency and any of the items of work
engagement. Moreover, sleep latency was not found to moderate the relation between momentary job demands and the items of work engagement. Randomization of slope was also performed for this analysis to see if there was a variance in momentary job demands on a daily level. However, no day-level variance was found for observed sleep latency. To summarize, the results showed that hypothesis 2 with observed sleep latency as moderator is rejected for all items of work engagement (DED1, DED2, VIG1 and VIG2). The results of all other analyses performed for this hypothesis can be found in Appendix 7.2.
Observed sleep efficiency
The analysis showed that more sleep efficiency led to less dedication (DED2) (g = -1.35, se = .46, p = .004) and less vigor (VIG1) (g = -.88, se = .40, p = .028). However, sleep efficiency did not moderate any of the relationships between momentary job demands and the items of work engagement.
Randomization of slope was performed for momentary job demands to find out if there was a within-person day-level effect of momentary job demands on the items of momentary work engagement. The results were similar to the analysis without randomization for DED2 (g = -1.40, se = .49, p = .004). The slope variance showed that there was a day-level effect of observed sleep efficiency on dedication (DED2) (s2 = .03, se = .01 p = .012). Furthermore, the analysis showed that there was a relation between
observed sleep efficiency and VIG1 (g = -.84, se =.40, p = .035) when job demands were randomized over days (s2 = .02, se =.01 p = .046).
The simple slopes test revealed that for both low (t = 2.85, p = .004) and high (t = 3.88, p < .001) levels of observed sleep efficiency, DED2 increased as momentary job demands also increased. No effect was found for either low (t = -.84, p = .401) or high (t = .05, p = .963) values of observed sleep efficiency on VIG1 in the simple slopes test. The results of the analysis performed for this hypothesis can be found in Appendix 7.2. Based on the results, the second hypothesis with observed sleep efficiency as a moderator was rejected for all items of work engagement (DED1, DED2, VIG1 and VIG2), as even the direction of the found effect on DED2 was the opposite of what was expected. Thus, no support has been found for hypothesis 2.
Hypothesis 3: Momentary job autonomy positively affects momentary work