15
Independence of observations
The World Values Survey guarantees a uniformly coordinated procedure for sampling and data
collection methods in all countries, including Singapore. Samples are randomly drawn under specified
conditions to retrieve a representative sample from the population. The data is collected by
professional organizations through face-to-face interviews, so it is almost impossible that one
respondent could influence another. The agencies have to report directly to the World Values Survey
and the country can only be included in a wave when all documentation is provided under very strict
conditions. Considering the extensive sampling and data collection procedure of the World Values
Survey, it would be safe to assume that this assumption is met.
Normal distribution
The normality assumption was investigated with the Shapiro-Wilk test, skewness and kurtosis values,
and histograms. The Shapiro-Wilk test was performed with a significance level of p > .05. The Null
hypothesis was that the filial piety scores were normally distributed for each category of the
independent variables. The results of the Shapiro-Wilk test of Normality (see Table 3 in the Appendix)
revealed that most values were significant, which indicated non-normality. The skewness values were
used to assess asymmetry and kurtosis was used to measure peaks in the distribution (see Table 4 in
the Appendix). The skewness (.43) and kurtosis (.84) values of filial piety were found within the
preferred threshold of -1 and 1, so the normality assumption was met. Most levels complied with the
acceptable range of -2 to 2 as well, except for primary education within the education levels and
unemployed within the employment statuses. The histograms offered a better view on the distribution
and deviating values (see Histrogram 1 and 2 in the Appendix). Both sublevels showed a positively
skewed, heavy leptokurtic distribution, so there were mainly high peaks in the left half of the
distributions. Considering that the One-way ANOVA is quite robust for violations of the normality
assumption and the histograms appear roughly normal, the choice was made to continue with the
One-way ANOVA.
Homogeneity of variances
The results of the Levene’s Test for Homogeneity of Variances are presented in the table below.
Variable
Levene Statistic
df1
df2
Sig.
Age
5.976
2
1061
0.003
Education level
0.763
2
1071
0.467
Employment status
1.125
2
1080
0.325
16
The assumption is met if p > .05. The results indicate that the assumption is met for education level
and employment status, so the variance in the scores on filial piety is the same for each sublevel of
education and each sub status of employment. However, the assumption was not met for age with p
< .01. In response, the Robust Tests of Equality of Means were consulted. The Welch test was found
statistically significant with p < .01, implying that there exists a difference between at least two of the
age groups regarding their mean scores on filial piety.
AGE
There were 1064 respondents who answered the question regarding their age, among whom
the average age was 43 years (M = 43.06, SD = 16.68), ranging 18-89 years. The variable was divided
into three groups: 18-37 years (N = 427), 38-57 years (N = 396), 58+ years (N = 241). The graph below
illustrates the distribution of scores on filial piety and shows how a specific age group scored compared
to the other groups.
A One-way ANOVA test was performed to investigate the differences between the age groups
regarding their mean score on filial piety. The outcomes showed that the effect of age on filial piety
was statistically significant with F (2, 1061) = 8.31, p < .01. The outcome implies that a statistically
significant difference was found between the means of at least two groups. Despite the outcome, the
actual differences in mean scores between the age groups were very small with an effect size of .02
using η². Because the results of the Levene’s test were statistically significant for p < .05, a Games-
Howell Post Hoc test was used to compare the age groups. The Games-Howell test revealed that there
was a statistically significant difference in the score on filial piety between the age groups 18-37 years
and 58+ years (p = .02), as well as between 38-57 years and 58+ years (p < .01). However, no statistically
0
20
40
60
80
100
120
140
160
1
1,25 1,5 1,75
2
2,25 2,5 2,75
3
3,25 3,5 3,75
4
4,25 4,5 4,75
5
5,25 5,5 5,75
6
Distribution filial piety scores for age groups
18-37
38-57
58+
17
significant difference was found between the age groups of 18-37 years and 38-57 years. The mean
scores of each group are presented in the graph below.
The graph shows that the age group 38-57 years predicts the highest scores on filial piety. In
line with the results of the One-way ANOVA, the difference between the mean scores of the age group
18-37 years and the age group 38-57 years was not enough to be significant. However, the difference
between the 58+ group and each of the younger age groups was indeed enough to be significant. This
implies growing older between 18-57 years would not predict statistically significant different mean
scores on filial piety, but growing beyond 57 years would lead to a statistically significant drop in the
mean score on filial piety. In conclusion, the results suggest that age can predict the score on filial
piety.
EDUCATION LEVEL
There were 1074 respondents who answered the question regarding their highest education
level attained, among whom the median level was ‘’incomplete secondary school: university-
preparatory type’’, ranging no formal education – university level degree. The variable was divided into
three groups: no/primary education (N = 205), secondary education (N = 579), university level
education (N = 290). The graph below illustrates the distribution of mean scores on filial piety and
shows how each education level scored compared to the other levels.
2,84
2,94
2,68
2
2,1
2,2
2,3
2,4
2,5
2,6
2,7
2,8
2,9
3
18-37 years
38-57 years
58+ years
Means Plot of Filial piety score for Age groups
Mean score