Assignment

Word count: 1500

Research question and hypothesis
In the following part the relation between stress, need for recovery and job satisfaction
is analysed. Stress is the independent variable, need for recovery the dependent variable and
the relation is moderated by the variable job satisfaction. In other words the hypothesis is that
there is a positive relation between stress and the need for recovery and this relation is
moderated by the job satisfaction. So job satisfaction moderates the strength of this relation. If
someone is highly stressed it leads to a high need for recovery. But if he is very satisfied with
his job, stress may affect his need for recovery less. And the other way around, e.g. the stress
affects him a lot more if he is dissatisfied with his job.

H1a: There is a positive relationship between stress and need for recovery.
H1b: There is a positive relationship between stress and need for recovery and this relationship
is moderated by job satisfaction, so that this relationship is weaker for higher values of job
satisfaction. Job Satisfaction

Need for
Stress Recovery
Figure 1: Conceptual framework

Analytical strategy
First of all, there has been a check of frequencies to examine if there were any errors in
the data. There were no errors found. Then there was dealt with missing values by excluding
cases listwise. This means that only cases that had no missing data in any variable were
analyzed. To assure there were no counter-indicative items, items that are phrased so that an
agreement with the item represents a low level of the construct being measured were recoded
into different variables. This means that rStress2 and rStress4, rJobsatisfaction2 and
rNeedforRecovery4 have been recoded and now represent Stress2, Stress4, Jsat2 and NR4.
After this, descriptive statistics, skewness, kurtosis and normality tests were
performed. The items of Stress and Job Satisfaction had no normally distributed items. All
items of stress have a skewness between -0.5 and 0.5, but the kurtosis is between -0.5 and -1,
which indicates the distribution might be a bit flatter. For JobSatisfaction1 skewness is between
-0.5 and -1 and the kurtosis between 0.5 and 1 which indicates that the distribution is
moderately negative and is more pointy and therefore has more scores in the tails. For
JobSatisfaction2 skewness is between -0.5 and -1 and the kurtosis is near zero, this indicates
that the distribution is moderately negative. For JobSatisfaction3 skewness is between -1 and -

1
2 and kurtosis is between 1 and 2 which indicates that the distribution is substantially negative
and has many scores in the tails and is pointy. Only NeedforRecovery11 was normally
distributed (skewness = .461 and kurtosis = -.404). For NeedforRecovery1 skewness was
between 0.5 and 1 and kurtosis between -0.5 and -1 which indicates that the distribution was
moderately positive and flatter. For NeedforRecovery2,3,5,9 skewness was between -0.5 and
0.5, but the kurtosis is between -0.5 and -1 which indicates that the distribution is a bit more
pointy, but close to being a normal distribution. For NeedforRecovery4,6,7,10 the skewness is
between 0.5 and 1 and the kurtosis is between -0.5 and 0.5 which indicates that the distribution
is moderately positive. NeedforRecovery8 has a skewness between -0.5 and 0.5, but a kurtosis
between -2 and -1 which indicates that that the distribution is flatter than normal. But “with
reasonably large samples, skewness will not make a substantive difference in the analysis”
(Tabachnick & Fidell, 2001, p. 74). “Kurtosis can result in an underestimate of the variance,
but this risk is also reduced with a large sample” (200+ cases: Tabachnick & Fidell, 2001, p.
75). In this research there were more than 200 participants, therefore the risk is reduced and
skewness would not make a substantive difference in the analysis.
There has been done a reliability analyses of all items of the variables Stress, Need for
Recovery and Job Satisfaction. Cronbach’s alpha of Stress, NeedforRecovery and
JobSatisfaction are all above 0.7 (Stress α=0.834; NeedforRecovery α= 0.891; Job Satisfaction
α= 0.824) which means the scale is good. The Corrected Item-Total Correlation of all items of
Stress, NeedforRecovery and JobSatisfaction are above 0.3, therefore all items are good. The
difference for Cronbach’s alpha if the item is deleted is only for Stress3 and JobSatisfaction1,3
lower than 0.1. Stress,1,2,4, JobSatisfaction2 and all items of NeedforRecovery have a higher
difference than 0.10 if the item is deleted. Scale means have been computed for Stress,
NeedforRecovery and JobSatisfaction, and were coded as StressTOT, JsatTOT and NRTOT
(table 1).

Table 1: Means, Standard Deviations, Correlations

Variables Mean SD 1 2 3 4 5 6
1. Gender 1.57 0.5 -
2. Age 15.67 11.44 -.11 -
3. Education 3.98 1.17 .01 -.20** -
4. Stress 3 0.86 .03 -0.02 .06 (.83)
5. Job Satisfaction 3.87 0.83 -.12 .141* -.01 -.14*(.82)
6. Need for -
Recovery 2.57 0.72 .04 -.13 .19** .49** .37** (.89)
** Correlation is significant at the 0.01 level (2-tailed)
* Correlation is significant at the 0.05 level (2-tailed)

2
Hypothesis testing - Hierarchical regression
Next a hierarchical regression analysis is used to examine the relation between the
independent variable (Stress) and the dependent variable (Need for Recovery). A hierarchical
multiple regression was performed to research the ability of Stress to predict Need for
Recovery, after controlling for gender, age and education. As a first step of the hierarchical
multiple regression, three predictors were entered: gender, age and education. This was done
for these three demographic variables so that a shared variability of these variables with the
predictive variable Stress can be controlled. Thus the observed effect of Stress on Need For
Recovery is independent of the effect of these three demographic variables.

Table 2: Hierarchical Regression Model of Need for Recovery

R2
R R2 Change B SE β T
Step 1 .230 .053** .053
Gender .055 .098 .038 .566
Age -.005 .004 -.079 -1.155
Education .122 .043 .195** 2.871
Step 2 .534 .285*** .232***
Gender .023 .085 .016 .269
Age -005 .004 -.079 -1.324
Education .102 .037 .163** 2.741
Stress .403 .048 .483*** 8.310
Statistical significance: *p <.05; **p <.01; ***p <.001

In step 1 the model was statistically significant F(3, 214) = 3.986; p <.01 and explained 5.3 %
variance in Need for Recovery. So it can be concluded that most variance of the variable Need
for Recovery is explained by other factors. In step 2 Stress was entered as a predictor and the
total variance explained by the model as a whole was 28.5% F(4, 213) = 21,202; p <0.001.
Thus the introduction of Stress explained an additional 23.2% in Need for Recovery, after
controlling for gender, age and education (R2 Change = .232; F(1,213) = 69.05; p < 0.001). In
the final model two out of four predictor variables were statistically significant, with education
recording a lower Beta value (β = .163, p <0.01) than Stress (β = .483, p <0.001). So although
education explains a small part of the variance, Stress has out of these 4 variables by far the
most predictive power on Need for Recovery.

3
Hypothesis testing – Testing the moderating effect
Finally the moderating of effect of Job Satisfaction was analyzed. Table 3 shows
relevant results of the analysis. By adding Job Satisfaction as a moderator the model is able to
explain 35,66% of the variance in Need for Recovery. This is an additional 7,16% compared to
the previous model. Further it could be shown that for a level of confidence of 95% there is a
significant (P=0.0176<0.05) negative moderating effect of Stress (β = -.1178, p <0.05). on the
relation between Stress and Need for recovery. The interaction leads to an increase of R² by
0.0166. This moderating effect is significant for all levels of Job Satisfaction (P<0.001). Low,
medium and high levels of Job Satisfaction have a significant moderating effect on the
relationship between Stress and Need for Recovery.

Table 3 Analysis of the moderating Effect of Job Satisfaction

Model Summary
R R-sq F df1 df2 P
.05972 .3566 41.0187 3.0000 222.0000 .0000

R-square increase due to interaction(s):

R2-chng F df1 df2 P
int_1 0.0166 5.7239 1.0000 222.0000 .0176

Conditional effects of X on Y at values of the moderator(s):

Job Satisfaction Effect SE T P
3.0463 .4639 .0580 8.0029 .0000
3.8717 .3667 .0457 8.0204 .0000
4.6971 .2694 .0642 4.1948 .0000

Table 4: Interaction Variables

Need for recovery

Variable B SE B t
Interaction Variables
Job Satisfaction .0823 .6180 1.8394
Stress .8229 .1915 .5270***
Job Satisfaction x Stress -.1178 .0493 -2.3925*
Statistical significance: *p <.05; ***p <.001

4
Conclusion
The hypotheses could be proven. The hierarchical regression analysis showed a
statistical relevant positive relation between the independent variable Stress and the dependent
variable Need for recovery. Without incorporating the moderating effect of Job Satisfaction the
model could explain 28.5% of the total variance of Need for recovery whereof 23.2% were
explained by Stress. By including Job Satisfaction as a moderator the overall variance that
could be explained rose to 35,66%. It was also shown that at all values of Job Satisfaction there
is a statistical significant moderating effect.

Grade is Passed

5
Reference

Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics.