Professional Documents
Culture Documents
Hareem Kapadia, Aamna Sami, Hira Asif | Methods of Data Analysis | April 22, 2018
Table of Contents
S Title Page
1 Abstract 2
2 Sample 3
3 Smoking between the sexes 8
4 Stress between the sexes 10
5 Stress levels across ages 14
6 Stress levels for different sources of stress: 19
Graphical representations
7 Stress levels for different sources of stress: An 22
analytical investigation
8 Stress levels across marital status groups 31
9 Stress levels of people with children 35
10 Internal factors affecting stress 38
11 Factors affecting life satisfaction 51
12 Life satisfaction of various education levels 69
13 Optimism across ages 78
14 Conclusion 84
PAGE 1
Abstract
The situations and pressures that cause stress are known as stressors. We usually think
of stressors as exogenous variables such as university, work, children or even midlife
crisis. The following study examines whether stress is caused by such external factors
or whether it is innate factors such as human nature that cause it. The study shows
that to a significant extent stress can be internal or self-generated especially through
negative experiences in life.
Another common misperception is that people smoke due to stress is also been
debunked. The study also examines the effect of age on optimism and finds that age
does influence optimism as one would expect. it also models the innate factors on life
satisfaction to check the extent to which they contribute to this variable.
PAGE 2
SAMPLE
Method
The study uses a self-administered questionnaire given to 439 people of different ages and
all backgrounds.
DEMOGRAPHICS:
I. AGE:
age 5 groups
PAGE 3
II. GENDER:
Sex
PAGE 4
III. MARITAL STATUS:
marital status
PAGE 5
SEPARATED 10 2.3 2.3 92.9
IV. EDUCATION:
Valid PRIMARY 2 .5 .5 .5
PAGE 6
SOME SECONDARY 53 12.1 12.1 12.5
PAGE 7
DOES GENDER AFFECT SMOKING BEHAVIOUR?
We begin by examining whether the mean number of smokers in males and females are
equal.
H0: The data for smokers in all categories of gender is normally distributed.
H1: The data for smokers in all categories of gender is not normally distributed.
Tests of Normality
Since significance is less than 0.05 in all cases of K-S and S-W tests, therefore, we
can conclude that the data for smokers of both genders is not normally
distributed.
Since our assumption does not hold we will use non-parametric test.
PAGE 8
MANN-WHITNEY U TEST:
We are now using M-W test to test our hypothesis:
H0: The mean number of male smokers = The mean number of female smokers.
H1: The mean number of male smokers ≠ The mean number of female smokers.
From the above table we can see that the significance is 0.483 (which is greater than
0.05), thus we will retain our null hypothesis.
We can conclude that the number of smokers among all the categories of gender
is equal.
PAGE 9
DOES GENDER CONTRIBUTE TO STRESS LEVEL?
Since significance level for K-S test (0.015 and 0.015) is less than 0.05 and significance level
for S-W test (0.096 and 0.176) is greater than 0.05, there is a 50% probability that the data
follows normal distribution.
Tests of Normality
Now we will examine if the mean stress level is equal among all categories of gender.
We will assume that normality is being supported and use a parametric test.
PAGE 10
This conclusion is supported by the QQ Plot above which shows data follows the line of
normality.
This conclusion is supported by the QQ Plot above which shows data follows the line of
normality.
PAGE 11
We will use F-test to check for equality of variances.
Group Statistics
PAGE 12
Independent Samples Test
Lower Upper
Now we need to examine if there is statistical evidence to prove stress levels for both
females and males are equal. For this purpose, we will use t-test.
H0: mean stress level for females = mean stress level for males
H1: mean stress level for females ≠ mean stress level for males
Alpha=0.025 t(0.025,416)= ±2.364
The critical value of t is ±2.364.
Since our t calculated (-2.948) < -2.364, therefore we reject H0.
We can conclude that the mean stress level of females and males is not equal.
PAGE 13
DOES AGE CONTRIBUTE TO STRESS LEVEL
Now we will check whether the mean stress level among all age groups is equal or not.
First, we need to check if the assumption for normality holds:
H0 : The data for stress levels among all age groups is normally distributed.
H1 : The data for stress levels among all age groups is not normally distributed.
Since significance level for K-S test (0.082, 0.200, 0.200) and S-W test (0.172, 0.483,
0.309, 0.388, 0.605) is greater than 0.05, therefore, we can conclude that the data
follows normal distribution.
Tests of Normality
PAGE 14
This conclusion is supported by the QQ Plot above which shows only slight deviation of
data from the line of normality in values ranging from 10-20 and 40-50.
This conclusion is supported by the QQ Plot above which shows only slight deviation of
data from the line of normality in values ranging from 30-40.
PAGE 15
This conclusion is supported by the QQ Plot above which shows only slight deviation of
data from the line of normality in values ranging from 10-20 and 30-40.
This conclusion is supported by the QQ Plot above which shows only slight deviation of
data from the line of normality in values ranging from 10-20 and 30-40.
PAGE 16
This conclusion is supported by the QQ Plot above which shows only slight deviation of
data from the line of normality in values ranging from 10-20 and 40-42.
Since the statistic equals 0.254 which is greater than 0.05, we conclude that
variances are not equal.
We proceed with the ANOVA test.
ANOVA
PAGE 17
We can see that the significance value is 0.005, which is below 0.05 and, therefore,
We can conclude that there is a statistically significant difference in the mean stress
levels between the different age groups.
Total perceived stress
Tukey HSDa,b
1 2
25 - 32 86 25.65
50+ 77 25.75
41 - 49 95 26.62 26.62
33 - 40 82 26.77 26.77
18 - 24 93 28.60
A Tukey post hoc test revealed that the stress levels were statistically significantly lower
in age groups 25-32 and 50+, and higher in age group 18-24. There was no statistically
significant difference in the stress levels of age groups 41-49 and 33-40.
PAGE 18
DOES THE SOURCE OF STRESS CONTRIBUTE TO STRESS LEVEL?
GRAPICAL REPRESENTATIONS
Now we take a look at the different sources of stress, and how many people contribute
their overall stress to these sources.
Out of the 439 people surveyed:
223 people reported that their stress was due to work.
12 people reported that their stress was due to spouse or partner.
12 people reported that their stress was due to relationships.
25 people reported that their stress was due to children.
27 people reported that their stress was due to family.
20 people reported that their stress was due to health/illness.
33 people reported that their stress was due to life in general.
54 people reported that their stress was due to money/finances.
16 people reported that their stress was due to time.
source of stress
PAGE 19
The box plot below shows peoples stress scored divided on the basis of the sources
of stress.
Work, family and life in general appear to have similar medians. Health/illness
and spouse or partner have the highest median stress levels. Time has the lowest
median stress level.
Work has three outliers. Relationships has one outlier. Life in general has one
outlier. Money/finances has two outliers.
PAGE 20
PAGE 21
DOES THE SOURCE OF STRESS CONTRIBUTE TO STRESS LEVELS?
ANALYTICAL INVESTIGATION
The next thing we chose to explore whether any group experiences more stress than
others based on the cause of stress. To check whether it is possible to run an ANOVA in
this case.
H1: Stress levels across stress factors are not normally distributed
L.O.S:
α= 0.05
Decision:
Reject H0 if sig < 0.05. Since majority of the sig>0.05, we do not reject H0.
Conclusion:
Tests of Normality
PAGE 22
CHILDREN .086 24 .200* .973 24 .752
TIME (lack of time, too much to do) .160 16 .200* .960 16 .656
This conclusion is supported by the QQ Plot above which shows only slight deviation of
data from the line of normality.
PAGE 23
This conclusion is supported by the QQ Plot above which shows deviation of data from
the line of normality in values ranging from 25-30, 30-35, and 40-45.
This conclusion is supported by the QQ Plot above which shows deviation of data from
the line of normality in values ranging from 25-30 and 35-40.
PAGE 24
This conclusion is supported by the QQ Plot above which shows only slight deviation of
data from the line of normality in values ranging from 20-25 and 35-40.
This conclusion is supported by the QQ Plot above which shows only slight deviation of
data from the line of normality in values ranging from 22-28 and 30-40.
PAGE 25
This conclusion is supported by the QQ Plot above which shows only slight deviation of
data from the line of normality in values ranging from 10-20 and 35-45.
This conclusion is supported by the QQ Plot above which shows only slight deviation of
data from the line of normality in values ranging from 10-20 and 35-45.
PAGE 26
This conclusion is supported by the QQ Plot above which shows only slight deviation of
data from the line of normality in values ranging from 10-20, 25-28 and 30-35.
This conclusion is supported by the QQ Plot above which shows deviation of data from
the line of normality in values ranging from 20-25 and 35-40.
PAGE 27
Therefore, run a parametric test.
We check the significance associated to the Levene’s statistic to see if the variances are
equal. Since the sig equals 0.553 which is > 0.05, we conclude that the variances are
equal.
PAGE 28
One-Way-ANOVA
HO: The mean stress levels from all sources of stress are equal
L.O.S:
α= 0.05;
df1= 8
df2= 408
Test Statistic:
F= S2between / S2within
Fcal= 1.898
F (0.05,8,408)= 2.02
Decision:
Conclusion:
The mean stress levels across the different sources of stress are equal.
ANOVA
PAGE 29
The Tukey HSD test shows a difference in mean stress levels for the people stressed out by
their spouses or partners and stressed out due to time. However, the F-statistic contradicts
this result. This may be due to a Type I error caused because of unequal and a few very
small sample sizes.
Tukey HSDa,b
1 2
b. The group sizes are unequal. The harmonic mean of the group
sizes is used. Type I error levels are not guaranteed.
PAGE 30
DOES MARITAL STATUS GROUP AFFECT STRESS?
The next thing we explore is whether people with various marital status experience more
stress than others using One-Way-ANOVA. To check whether it is possible to run an
ANOVA in this case, we first need to check for normality.
H0: Stress levels across marital status groups are normally distributed
H1: Stress levels across marital status groups are not normally distributed
L.O.S:
α= 0.05
Decision:
Reject H0 if sig < 0.05. Since majority of the sig>0.05, we do not reject H0.
Conclusion:
PAGE 31
Tests of Normality
We check the significance associated to the Levene’s statistic to see if the variances are
equal. Since the significance equals 0.553, which is > 0.05, we conclude that the variances
are equal.
PAGE 32
One-Way-Anova
HO: The mean stress levels across all marital status groups are equal
H1: The mean stress level for atleast one marital status group differs
L.O.S:
α= 0.05;
df1= 7
df2= 425
Test Statistic:
F= S2between / S2within
Fcal= 2.749
F (0.05,7,425)= 2.09
Decision:
Conclusion:
The mean stress level for at least one marital status group differs.
ANOVA
PAGE 33
Tukey HSDa,b
1 2
WIDOWED 7 22.43
SEPARATED 10 30.90
b. The group sizes are unequal. The harmonic mean of the group
sizes is used. Type I error levels are not guaranteed.
The Tukey HSD table confirms our finding which shows that the means stress levels of
people that are married are higher and the mean stress levels of people that are widowed
are lesser than the rest of the marital status groups.
PAGE 34
DO CHILDREN AFFECT STRESS LEVELS?
In our sample, we have 183 people have children and 249 that do not have children. We
hypothesize that people having children may have a higher stress score than those not
having children for which we run and independent samples t-test.
Group Statistics
L.O.S
α= 0.05; α/2 = 0.025
df1= 182; df2= 249
Test statistic:
F= (S2children/S2no children)*(σ2no children / σ2children)
Decision:
Since Fcal=2.832, we reject H0.
Conclusion: PAGE 35
Variances of the two groups are unequal.
We therefore use a non-pooled t-test.
L.O.S:
α= 0.05
α/2= 0.025
df= 415.5
Test Statistic:
t = ((x1 bar- x2 bar) – (mew1-mew2))/Root((s1^2/n1)+(s2^2/n2))
Decision:
Since tcal=1.249, we do not reject H0.
Conclusion:
The mean stress to those people having children and those having no children is equal.
Lower Upper
PAGE 36
Equal 2.832 .093 - 430 .221 -.697 .569 - .421
variances 1.225 1.816
Total assumed
PAGE 37
INTERNAL FACTORS AFFECTING STRESS
Now that we have thoroughly analyzed the external factors causing stress and have
concluded that whether a person has or doesn’t have children does not effect stress score,
nor does marital status, or age, we turn our attention to the endogenous factors that we
scored using respondents answers from the questionnaire.
Testing of Beta
L.O.S:
α= 0.05
Test Statistic:
t= (b2-β2)/ Sb2
tcal= -11.005
t(0.05, 438)=-1.28
Decision:
Conclusion:
PAGE 38
Coefficientsa
L.O.S:
α= 0.05
Test Statistic:
t= (b2-β2)/ Sb2
tcal= -16.051
Decision:
Conclusion:
PAGE 39
Mastery score is a useful predictor of stress.
Coefficientsa
H0: β2=0; Positive effect score is not a useful predictor of stress score.
L.O.S:
α= 0.05
Test Statistic:
t= (b2-β2)/ Sb2
tcal= -10.219
Decision:
Conclusion: PAGE 40
H0: β2=0; Negative effect score is not a useful predictor of stress score.
H1: β2>0; Negative effect score is not a useful predictor of stress score.
L.O.S:
α= 0.05
Test Statistic:
t= (b2-β2)/ Sb2
tcal= 18.944
Decision:
Conclusion:
H0: β2=0; Life satisfaction score is not a useful predictor of stress score.
L.O.S:
α= 0.05
Test Statistic:
t= (b2-β2)/ Sb2
tcal= -11.808
Decision:
Conclusion:
PAGE 42
Life satisfaction score is a useful predictor of stress.
Coefficientsa
PAGE 43
Perceived Control of Well being as a predictor of Stress
H0: β2=0; Perceived control & well being score is not a useful predictor of stress score.
H1: β2<0; Perceived control & well being score is a useful predictor of stress score.
L.O.S:
α= 0.05
Test Statistic:
t= (b2-β2)/ Sb2
tcal= -14.683
Decision:
Conclusion:
PAGE 44
Regression Model
After running the useful predictor tests, we can conclude that optimism (D1), mastery
(D2), positive affect (D3), negative affect (D4), life satisfaction (D5) and perceived control
and well being (D6) scores are all useful predictors of stress score.
We can therefore generate a linear regression model to predict a persons stress score
based on these factors:
Coefficientsa
PAGE 45
Model Summaryb
The fitted regression model explains a 63.1% variation in stress due to variation in the
following factors. The value of R shows a strong positive relationship between stress and
independent variables. A 10% increase in the independent variables causes a 9.94%
increase in stress.
The model shows us that optimism, mastery, positive affect (characteristics), life
satisfaction and perceived control and well being are all inversely related to stress.
A one-unit increase in a persons self scored optimism decreases their stress score by 0.11
units.
A one-unit increase in a persons mastery score decreases their stress score by 0.331 units.
This is to say that the higher a people perceives themself to be in control of life, the less
stress he/she feels.
A one-unit increase in a person’s life satisfaction score decreases stress score by 0.154
units.
A one-unit increase in one’s perceived control and well-being (control over ones well-
being) decreases stress score by 0.078 units.
The constant 37.828 shows that stress mainly depends on other variables that have not
been accounted for in our model.
PAGE 46
Goodness of fit of the model
L.O.S
α = 0.05
Test statistic:
F= S2between / S2within
F(0.05,6,417)= 2.18
Fcal = 118.897
Decision:
Conclusion:
ANOVAa
PAGE 47
Regression 9127.201 6 1521.200 118.897 .000b
b. Predictors: (Constant), Total percieved control and well being, Total life satisfaction, Total
negative affect, Total positive affect, Total Optimism, Total Mastery
I. Normally distributed.
The table and scatter plot below shows that residuals are approximately normally
distributed.
Tests of Normality
Kolmogorov-Smirnova Shapiro-Wilk
PAGE 48
II. Independence of residuals
This graph shows us that the residuals are following a linear trend which means residuals
are not independent.
PAGE 49
Therefore, we can conclude that our model is violating the assumptions of residuals.
PAGE 50
FACTORS AFFECTING LIFE SATISFACTION
First, the variables will be checked individually if they are useful predictors of life
satisfaction. We begin with testing whether the variables follow normality or not.
Optimism
Since significance associated to the KS test and SW test equals 0 which is < 0.05, we
reject H0
Tests of Normality
Kolmogorov-Smirnova Shapiro-Wilk
PAGE 51
This conclusion is supported by the QQ Plot above as it shows diversion of the data from
the line of normality in the values ranging from 5 to 15.
Now we run a regression between life satisfaction and optimism to check if optimism is a
useful predictor of life satisfaction.
Coefficientsa
Model Unstandardized Coefficients Standardized t Sig.
Coefficients
Since the significance associated to optimism equals 0 which is < 0.05, we reject H0,
thus concluding that optimism is indeed a useful predictor of life satisfaction.
PAGE 52
Positive Effect
Tests of Normality
Kolmogorov-Smirnova Shapiro-Wilk
Since significance associated to the KS test and SW test equals 0 which is < 0.05, we
reject H0.
Therefore, we can conclude that positive effect does not follow normal distribution
PAGE 53
This conclusion is supported by the QQ Plot above which shows deviation of data from
the line of normality in values ranging from 10 to 20 and 40 to 50.
Now we run a regression between life satisfaction and positive effect to determine
whether positive effect is a useful predictor of life satisfaction.
Coefficientsa
Since the significance associated to positive effect equals 0 which is < 0.05, we reject H0,
thus concluding that positive effect is indeed a useful predictor of life satisfaction.
PAGE 54
Negative Effect
Tests of Normality
Kolmogorov-Smirnova Shapiro-Wilk
Since significance associated to the KS test and SW test equals 0 which is < 0.05, we
reject H0.
Therefore, we can conclude that negative effect does not follow normal distribution.
This conclusion is supported by the above QQ Plot which shows deviation from the line
of normality in the overall data.
PAGE 55
Now we run a regression between life satisfaction and negative effect to determine
whether negative effect is a useful predictor of life satisfaction
Life Satisfaction
Coefficientsa
Model Unstandardized Coefficients Standardized t Sig.
Coefficients
Since the significance associated to negative effect equals 0 which is < 0.05, we reject H0,
thus concluding that negative effect is indeed a useful predictor of life satisfaction.
PAGE 56
Stress
Testing for normality
Tests of Normality
Kolmogorov-Smirnova Shapiro-Wilk
Since significance associated to the KS test and SW test equals 0 and 0.021 which are
both < 0.05, we reject H0.
Therefore, we can conclude that stress does not follow normal distribution.
This conclusion is supported by the above QQ Plot which shows a deviation from the line
of normality in the data ranging from 40 to 50.
Now we run a regression between life satisfaction and stress to determine whether stress
is a useful predictor of life satisfaction.
PAGE 57
Useful predictor test
Coefficientsa
Since the significance associated to stress equals 0 which is < 0.05, we reject H0, thus
concluding that stress is indeed a useful predictor of life satisfaction.
PAGE 58
Self-esteem
Tests of Normality
Kolmogorov-Smirnova Shapiro-Wilk
Since significance associated to the KS test and SW test equals 0 which is < 0.05, we reject
H0.
Therefore, we can conclude that self-esteem does not follow normal distribution.
This conclusion is supported by the above QQ Plot which shows deviation from the line
of normality in the overall data.
Now we run a regression between life satisfaction and self-esteem to check if self-esteem
is a useful predictor of life satisfaction.
PAGE 59
Useful predictor test
Coefficientsa
Since the significance associated to self-esteem equals 0 which is < 0.05, we reject H0, thus
concluding that self-esteem is indeed a useful predictor of life satisfaction.
PAGE 60
Social Desirability
Tests of Normality
Kolmogorov-Smirnova Shapiro-Wilk
Since significance associated to the KS test and SW test equals 0 which is < 0.05, we reject
H0.
Therefore, we can conclude that social desirability does not follow normal distribution.
This conclusion is supported by the above QQ Plot which shows deviation from the line
of normality for data ranging from 0 to 1.
Now we run a regression between life satisfaction and social desirability to check if
social desirability is a useful predictor of life satisfaction.
PAGE 61
Useful predictor test
Coefficientsa
Since the significance associated to social desirability equals 0.024 which is < 0.05, we
reject H0, thus concluding that social desirability is indeed a useful predictor of life
satisfaction.
PAGE 62
Total Control
Tests of Normality
Kolmogorov-Smirnova Shapiro-Wilk
Total percieved control and .058 430 .002 .987 430 .001
well being
Since significance associated to the KS test and SW test equals 0.002 and 0.001, respectively,
which are both < 0.05, we reject H0.
Therefore, we can conclude that total control does not follow normal distribution.
This conclusion is supported by the above QQ Plot which shows deviation from the line
of normality in the data ranging from 20 to 30 and 80 to 90.
PAGE 63
Now we run a regression between life satisfaction and total control to check if total
control is a useful predictor of life satisfaction.
Coefficientsa
Since the significance associated to total control equals 0 which is < 0.05, we reject H0, thus
concluding that total control is indeed a useful predictor of life satisfaction.
Now that we have determined which variables are useful predictors of life satisfaction,
we will now create a regression models with the useful predictors.
Model Summaryb
The fitted regression model explains a 38% variation in life satisfaction due to variation
in optimism, positive effect, negative effect, stress, self-esteem, social desirability and
total control. A low R2 in this case is not a negative indication of the model as this model
predicts human behavior which is harder to predict than another physical phenomenon.
The value of R shows a strong, positive relationship between life satisfaction and the
PAGE 64
independent variables. A 10% increase in the independent variables causes a 61.6% change in
life satisfaction.
ANOVAa
Now we check the significance associated to F. The significance equals 0 which is <0.05. This
indicates that R2 statistically significant and can be generalized for predictions
PAGE 65
Model Unstandardized Coefficients Standardized t Sig.
Coefficients
PAGE 66
Now we check the residual assumptions:
i. Residuals follow normality
Tests of Normality
Kolmogorov-Smirnova Shapiro-Wilk
Since significance associated to the KS test and SW test equals 0.052 and 0.236, respectively,
which are both > 0.05, we do not reject H0.
This conclusion is supported by the above QQ Plot which shows little to no deviation
from the line of normality.
PAGE 67
ii. Residuals are independent
R2linear = 0.62
R2quadratic = 0.622
R2cubic = 0.629
All fitted models have an R2 > 0.3 which shows that the residuals are not independent.
PAGE 68
DOES EDUCATION LEVELS EFFECT LIFE SATISFACTION?
Another test we will conduct will check if different education levels have different mean
life satisfaction scores.
Tests of Normality
PRIMARY .260 2 .
PAGE 69
Primary
For this study, primary education has been ignored due to negligible sample size (n=3).
Some secondary
Since significance associated to the KS test and SW test both equal 0.006 which is < 0.05, we reject
H0.
Therefore, we can conclude that the life satisfaction of secondary education does not follow normal
distribution.
This conclusion is supported by the above QQ Plot which shows deviation from the line
of normality in the data ranging from 8-14 and 32-37.
PAGE 70
High School Education
Since significance associated to the KS test and SW test equals 0.091 and 0.483,
respectively, and since both are > 0.05, we do not reject H0.
Therefore, we can conclude that the life satisfaction of high school education follows
normality.
This conclusion is supported by the above QQ Plot which shows little to no deviation
from the line of normality in the data.
PAGE 71
Additional Training
Since significance associated to the KS test and SW test equals 0.157 and 0.249,
respectively, and since both are > 0.05, we do not reject H0.
Therefore, we can conclude that the life satisfaction of additional training follows
normality.
This conclusion is supported by the above QQ Plot which shows little to no deviation
from the line of normality in the data.
PAGE 72
Undergraduate
Since significance associated to the KS test and SW test equals 0.038 and 0.015,
respectively, and since both are < 0.05, we reject H0.
Therefore, we can conclude that the life satisfaction of undergrad education does not
follow normality.
PAGE 73
Postgrad
Since significance associated to the KS test and SW test equals 0.200 and 0.776,
respectively, and since both are > 0.05, we do not reject H0.
Therefore, we can conclude that the life satisfaction of postgrad education follows
normality.
This conclusion is supported by the above QQ Plot which shows deviation from the line
of normality in the data ranging from 5-11.
PAGE 74
Since 3/5 follow normality, we go for a parametric test. In this case since there are more
than 2 populations, we go for One-way ANOVA.
We check the significance associated to the Levene statistic to see if the variances are
equal. Since the statistic equals 0.129 which is > 0.05, we conclude that variances are
equal.
One-Way Anova
ANOVA
Total life satisfaction
Now we carry out the F-test to check for our original hypothesis.
𝑆𝑆 𝑏/𝑤
α=0.05 Test Statistic: Fcal = 𝑆𝑆 𝑤𝑖𝑡ℎ𝑖𝑛
Critical Value: F(0.05 (5, 430)) = 2.23 Rule of rejection: Reject H0 if Fcal > 2.23
Decision Statement: Since Fcal < 2.23, we do not reject H0, hence concluding that the
mean life satisfaction of all education levels does not have a significant difference.
PAGE 75
Total life satisfaction
Tukey HSDa,b
PRIMARY 2 22.00
This conclusion is supported by the Tukey table above which shows that all means lie in
the same column, meaning there is no significant difference between them.
PAGE 76
ARE YOUNGER PEOPLE MORE POSITIVE?
A popular belief is that when one is younger, he/she has a more positive and optimistic
outlook on life. We will conduct a test to check levels of optimism throughout various
age groups.
First we check for normality in order to determine whether to use a parametric test or a
non-parametric test.
Tests of Normality
PAGE 77
Age group: 18-24 years
Since significance associated to the KS test and SW test equals 0.043 and 0.021,
respectively, and since both are < 0.05, we reject H0.
Therefore, we can conclude that the optimism of ages 18-24 does not follow normality
This conclusion is supported by the above QQ Plot which shows deviation from the line
of normality in the data ranging from 5-14.
PAGE 78
Age group: 25-32 years
Since significance associated to the KS test and SW test equals 0.005 and 0.241,
respectively, and since one is < 0.05, we reject H0.
Therefore, we can conclude that the optimism of ages 25-32 does not follow
normality.
This conclusion is supported by the above QQ Plot which shows deviation from the line
of normality in the data ranging from 10-15.
PAGE 79
Age Group: 33-40 years
Since significance associated to the KS test and SW test equals 0.008 and 0.005,
respectively, and since both are < 0.05, we reject H0.
Therefore, we can conclude that the optimism of ages 33-40 does not follow
normality.
This conclusion is supported by the above QQ Plot which shows deviation from the line
of normality in the data ranging from 10-15 and 28-30.
PAGE 80
Age Group: 41-49 years
Since significance associated to the KS test and SW test equals 0.06 and 0.029,
respectively, and since one is < 0.05, we reject H0.
Therefore, we can conclude that the optimism of ages 41-49 does not follow
normality.
This conclusion is supported by the above QQ Plot which shows deviation from the line
of normality in the data ranging from 6-16.
PAGE 81
Age group: 50 +
Since significance associated to the KS test and SW test equals 0.004 and 0.022,
respectively, and since both are < 0.05, we reject H0.
Therefore, we can conclude that the optimism of ages 50+ does not follow normality.
This conclusion is supported by the above QQ Plot which shows deviation from the line
of normality in the data ranging from 13-17.
PAGE 82
Since 5/5 do not follow normality, we go for non-parametric test, which is the Kruskal-
Wallis H independent samples test.
Test Statisticsa,b
Total Optimism
Chi-Square 15.206
df 4
Asymp. Sig. .004
Since significance equals 0.004 which is < 0.05, we reject H0, hence concluding that mean
optimism for all ages has a significant difference.
PAGE 83
Conclusion
After carefully examining the effect of the demographic/external variables, i.e. age, gender
and education levels and if a person has children on stress, we find that none of the four
factors are have a strong enough effect on stress. The data therefore negates various
assumptions that we may make in real life.
For example, a person may assume that a younger person has less worries and therefore
takes less stress in life or that a person without children probably experience less stress.
However, after a thorough investigation of our sample, the study suggests that no such
conclusions can be drawn.
After the analysis of the external variables, we draw our focus to the effects of internal
variables on stress. These include optimism, positive and negative experiences, how much
a person feels he is in control of his life and of his well- being and how satisfied a person is
with their life. The results are coherent with our assumptions. We find that optimism,
positive experiences, life satisfaction and perceived control of well-being all have an
inverse effect on stress, while negative experiences have a very significant positive effect.
In this model, optimism, positive effect, better self-esteem and better control in life have a
direct and positive effect on life satisfaction. However, it is surprising that negative
experiences also have a positive effect on life satisfaction. Not surprisingly, stress has a
negative effect.
The regression models do not appear to be not a good fit. However, since this is a study of
human behavior, there is always a huge element of unpredictability.
PAGE 84