Example 1

A statistical analysis report to evaluate a … … … intervention within a student

population.

Introduction

Statistics can be defined as the science of collecting, organising and analysing data
from a sample population to make inferences about the target population (Zulfiqar &
Bhaskar, 2016). Statistics are essential in public health as they provide evidence to
help identify health problems and causes of illness, as well as allowing health
professionals to prioritise health resources (World Health Organization, 2014). Bruce,
Pope and Stanistreet (2008) describe two types of statistics, descriptive and
inferential. Descriptive statistics are used to describe and summarise information
about a population, for example demographic information and patient health
information. Inferential statistics, on the other hand, allow researchers to make
inferences about the general population based on findings from the sample population
being investigated (Bruce, Pope & Stanistreet, 2008).

Both types of statistics may be used to help formulate a research question. To answer
the research question, a hypothesis needs to be formulated and tested.
…………………………………………………………………………………………………
…………………………………………………………………………………………………
…………………………………………………………………………………………………
…………………………………………………………………………………………………
…………………………………………………………………………………………………
…………………………………………………………………………………………………
…………………………………………………………….. (Ferrill, Brown & Kyle, 2010).

When the null hypothesis is wrongly rejected, it is referred to as a Type-I error, and
when it is wrongly accepted this refers to a Type-II error (Lehmkuhl, 1996). Type-II
errors occur due to low statistical power. Bruce, Pope and Stanistreet (2008) define

1
power as the probability of correctly rejecting the null hypothesis and recommend
setting a power of 80-90%. A key factor of statistical power is ensuring a large enough
sample size. ……………………………………………………………………………...

Parametric and non-parametric tests

Two types of statistical investigations were considered when answering questions in

this report: parametric and non-parametric tests. Lehmkuhl (1996) states that several
assumptions must be met before conducting parametric tests: samples are random,
experiments are repeatable, normal distribution of data, and samples have similar
variance. However, Field (2018) states that the dependent variable must also be
continuous, for example weight. Non-parametric tests, however, do not rely on the
above assumptions so are quicker and easier to perform (Lehmkuhl, 1996). However,
non-parametric tests are less powerful, so they should only ever be used when the
assumptions of non-parametric tests are not met (Pallant 2013).

Aim of this report

In this report relevant statistical analyses will be used to evaluate a healthy lifestyle
education intervention. The aim of the intervention was to promote healthy lifestyle
behaviours among university students. Improving health literacy (HL) and reducing
weight were two key objectives of the study. There were …… participants recruited
from two different towns, ……. were from town A and …… were from town B. Of the
total participants, ……… (x %) were male and ……. (y%) were female. The age of
students ranged between xxx and xxx, with a mean age of xxxxx. xxxxxxxxxxx (yy %)
were assigned to the active intervention group and xxxxxx (x %) were assigned to the
control group. A list of variables measured are included in appendix 1.

Preliminary analyses and investigations

2
Prior to analysing data, it essential to carry out data screening to identify and correct
any omissions or errors in the data set (Pallant, 2013). Descriptive statistics were run
for all variables measured.

Frequencies were performed for the all categorical variables and are presented in
output 1.1. There was one missing value for the location variable. There was no record
of this participant’s location therefore the missing value could not be corrected. There
were no other missing values and minimum and maximum values were within the
correct range for all variables.

Table 1.1 Categorical variables

Variable 1 Variable 2 Variable 3 Variable 4 Variable 5

N Valid
Missing
Minimum
Maximum

Descriptives were performed for all continuous variables and are presented in output
1.2. There were no missing variables and all minimum, maximum and mean values
were within expected ranges.

Table 1.2 continuous variables

N Minimum Maximum Mean Std. Deviation
Variable 1
Variable 2
Variable 3
Variable 4
Valid N

3
Question x: Determine whether there are ……………………………………..?

Choice of statistical test

Xu et al. (2017) state that when the outcome variable is continuous and there are two
independent samples, an independent t-test should be selected to measure if there is
a significant difference in the mean values. Pallant (2013), however, states that the
independent variable should be categorical. Assumptions of normal distribution and
similar variance across the two groups must be checked prior to conducting an
independent t-test.

Preliminary investigations

Normal distribution

…. scores were checked in both groups to detect if distribution was normal. Ghasemi
and Zahediasal (2012) state that normal distribution can be checked visually or via
tests that check if the sample distribution differs significantly from the normal
distribution. A histogram is commonly used to visually check for normal distribution,
which is identifiable through a bell-shaped curve. Figure 1 and figure 2 demonstrate
normal distribution of HL scores in both the control and intervention group.

4
 Figure 1. Distribution of … Scores (this is a fictitious diagram the actual figure is removed to minimise potential
for inadvertent plagiarism)

Figure 2. Distribution of …Scores (this is a fictitious diagram figure is removed to minimise potential for
inadvertent plagiarism)

Ghasemi and Zahediasal (2012) argue that visually checking for normal distribution
can be unreliable and advise using statistical tests. A common test for checking the
this is a Kolmogorov-Smirnov (K-S) test.

5
A K-S test was carried out to assess if the xxx scores were normally distributed in the
control and intervention groups. The xxx scores in the control group, D(xx) = xxx, p =
xxx, and the intervention group, D(xx) = xxx, p = xxx, were both significantly non-
normal.

A significant K-S value indicates that distribution is not normal (Field, 2018), which
suggests the xx scores were not normally distributed. This contradicts the visual
interpretation of the histograms. However, Ghasemi and Zahediasal (2012) argue that
the K-S test has low power and that the Shapiro-Wilk (S-W) test is a better option for
testing normality.

A S-W test was then carried out to assess if the xxx scores were normally distributed
in the control and intervention groups. The xxx scores in the control group, W(xx) =
xxx, p = xxx, and the intervention group, W(xx) = xxx, p = xxx, did not deviate
significantly from normal.

After considering the results from the histogram, K-S test and S-W test, it was
concluded that xxx scores were normally distributed.

Homogeneity of variance

A Levene’s test was carried out to assess if variance of … scores were similar across
the control and intervention groups. For … scores, the variances for the control and
intervention group were not significantly different, F(…) = xxx, p = xxx. Similar variance
was therefore assumed.

6
Results

An independent t-test was conducted to determine if there were any difference in mean
… scores between the intervention and control groups (see figure 3).

There was a significant difference in scores in the intervention group (M = xxx,

SD=xxx) and the control group (M = xxx, SD=xxx); t (xx) = xxx, p = 0.xxx, two-tailed.
The mean difference = xx, 95% CI: xxx to xxx.

Figure 3. Mean xxx Scores for the Intervention and Control Groups (Figure is removed to minimise potential
for inadvertent plagiarism)

Conclusion

The aim of this report was to evaluate a xxx intervention. Findings from statistical
analyses indicate that those in the intervention group had significantly higher xxx
scores and lost more xxx, than those in the control group. However, as the statistical
tests only reject the null hypotheses, no further conclusions can be made.

7
No association was found between xxx status and intervention group status. This
could potentially be explained by the random allocation of participants into each group.
It is also possible that the time-scale of the study would be too short to have any impact
on xxx prevalence.

The change in xxx could not be significantly predicted by xxx, xxx, or xxx scores.
However, it may be worth repeating this model with a larger, more diverse sample. As
all participants were students, the sample was young (age: mean of xx years and
median xx years). It is also possible that the samples mean xxx would be higher than
the average population. Also, males (xx%) were highly underrepresented in this
sample.

8
Reference

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

9
10