Elucidation of Statistics in Research Articles 2022

Elucidation of Statistics in Research Articles
1

Paper 1 Review
1. This paper under review consisted of two sets of hypotheses to inferentially analyze
the two objectives of the study. The null hypothesis and the alternative hypothesis for each of
the two objectives have been provided with corresponding independent variable, as well as
the dependent variables.
The First Objective:
Null hypothesis: There was no difference in levels of AOD used at “end of school
celebration” and other “peer-based social events”.
Alternate hypothesis: There was a significant difference in levels of AOD used at
“end of school celebration” and other “peer-based social events”.
Independent Variable: Event Type (school celebration, “peer-based social events”)
Dependent Variable: “Adolescent alcohol and other drugs" (AOD) use
The Second Objective:
Null hypothesis: There was no association between harm-minimization strategies and
harms experienced at the celebrations, controlling for other factors.
Alternate hypothesis: There was a significant association between harm-minimization
strategies and harms experienced at the celebrations, controlling for other factors.
Independent Variables: “Average quantity of alcohol” used in a day at any event,
drug use, “alcohol-related safety strategies”, gender, location of stay, and survey type
(online or offline).
Dependent Variable: Harms/Consequences experienced at the celebrations (17
categories; hangover, vomiting, unprotected sex, blackout, etc.).
2. Convenience sampling or opportunistic sampling strategy was used in the research
(Lam, Liang, Chikritzhs, & Allsop, 2014). A two-fold design (pre and post celebration) with
2

a self-report sampling methodology was adopted in the study. Participants were asked
questions on a face to face basis, and also using an online questionnaire. Likert scale and free
response form were used in both the surveys. Psychometrically validated items were
interleaved in the survey form. The pre-celebration survey was designed to collect quantity-
specific anticipations of “Adolescent alcohol and other drugs" (AOD) usage. Information on
the expectation of the lenient nature of the celebration was noted. Also, parental views on
alcohol and AOD usage at students' last social events were noted. The post-event survey
collected information about AOD use, especially about AOD use of peers and its negative
impacts. Also, data about harm reduction strategies were collected.
The primary advantage of this sampling methodology was that it was cost-effective,
since participants reported about all the factors and the researcher did not invest money to
measure or collect any type of information. Also, less time was spent on the data collection
procedure, and it helped in avoiding subsidiary costs.
Among disadvantages, the first and foremost drawback was the possibility of
misreported data, either exaggerated or understated in nature resulting in a highly skewed
distribution of data. Also, in case of collecting delayed collection of data after completion of
the last celebrations might create incidence where respondents forget the exact sequence of
events.
3. Information was collected from 541 young school leavers’ (56% female) in the pre-
celebration stage. The majority of the students were 17 years old (91%), and the rest were all
18 years (9%). The post-celebration information was gathered from 405 students (50%
female), where 94% of participants were 17 years old and the rest 6% were of 18 years. The
sample only consisted of young students aged between 17-18 years of age. The research was
3

based solely on assessing the impact of harm reduction strategies, and other factors on the
consequences faced in a celebration party.
4. Two inferential analyses were used in the study, and they have been discussed below.
“Wilcoxon signed-rank test” (non-parametric) was used to pair-wise compare usage
of “Adolescent alcohol and other drugs" AOD between last school leavers’ celebrations event
and at a normal celebration day.
The reason behind using the "Wilcoxon signed-rank test” was due to the particular
sampling strategy of the research. As the collected data could have generated a skewed
distribution for differences between pre and post scores, a parametric test was required for
pairwise comparison.
Logistic regression models were used to estimate the impact of “average quantity of
alcohol used in a day at any event”, drug use (other than alcohol), “alcohol-related safety
strategies”, location of stay, gender, and survey type (online or offline) on consequences or
harms experienced at the celebrations.
The reason behind using logistic regression modeling was the categorical nature of the
outcome variable (Mood, 2010). Seventeen different categories (hangover, vomiting,
unprotected sex, blackout, etc.) were present for the consequences or harms experienced at
the celebrations. The impact of the six independent factors (average quantity of alcohol used,
drug other than alcohol, alcohol-related safety strategies, gender, location of stay, and survey
type) on consequences faced at celebrations was assessed using logistic regression analyses.
5. In comparison with participants who were engaged in the greatest frequency of
protective strategies, odds ratio (OR) for engaging in unprotected sex was equal to 10.92 for
participants with the lowest frequency of protective strategies.
4

The OR implied that young people, who were not betrothed with safety strategies to
counter the consequences or harms experienced at the celebrations due to AOD, were almost
11 times more likely to engage in unprotected sex, because of AOD abuse. Here, the average
quantity of alcohol used at any event, drug use, gender, location of stay, and survey type was
controlled to estimate the individual impact of safety strategies on handling the consequences
experienced at the celebrations. Impact of safety strategies was estimated keeping other
factors as constant.
6. The sampling strategy of the study was opportunistic, and data was collected from any
willing young student of 17-18 years of age. Also, the sample data was collected from school
leavers’, and present students from schools were not considered for sampling. Generally,
school leavers’ have high alcohol consumption capabilities, and the same thing is not valid
for all the students in the school (Pettigrew et. Al., 2015). Generalization of the results for the
national population of schools was not a realistic proposition. It has to be also noted that the
sample size was very large and represented almost one-third of the celebrating population.
But, due to the homogeneity of sample respondents (school leavers), the generalization of the
results for the entire population of schools would be an amorphous step.
Paper 2 Review
1. The research objective of the study was to evaluate the pervasiveness of self-
reported diabetes (Wong, Leung, Tsang, Lo, & Griffiths, 2013). The data for the research
work was collected using “territory-wide household surveys”. The sample was collected
randomly and represented the entire Hong Kong population. Age, household income, and
gender were considered as the three predictors of the outcome variable (self-reported
5

diabetes) in binary logistic regression. The study also examined the factors which were found
to be independently related to diabetes. The null and alternative hypotheses of the study can
be presented as below.
Null hypothesis: H01: Increase in age for males and females had equal impact on
self-report of diabetes.
Alternative hypothesis: HA1: The increase in age for males had a significantly
different impact on the self-report of diabetes compared to that of the females.
Null hypothesis: H02: Monthly income (household) was not associated with self-
report of diabetes (referent income group earning above $ 50,000).
Alternative hypothesis: HA2: Monthly income (household) was significantly
associated with self-report of diabetes.
2. The demographic characteristics according to the statistics reported in the paper have
been reported underneath.
The investigators documented demographic information regarding age, sex and
monthly household income of participants.
Information from a total of 121,895 interviewed people was collected from the
surveys in 2001, 2002, 2005, and 2008. Among the total survey people, 103,367 were noted
to be adults (aged above or equal to 15 years) with an average age of 38.2 years. Among the
adult participants, the majority of people (18.2%) belonged to the age group of 35-44 years,
followed by 16.4% people aged between 45-54 years, 14.6% people between 25-34 years,
and 13.8% aged between 15-24 years. The rest of the 21.8% of people were aged above 55
years.
6

Gender wise analysis revealed the presence of 49.8% male, and 50.2% female
participants. Monthly household income of 42.4% of the interviewed people was between
10,000 and 24,999 (HK dollars). Among the rest, the income of 27.4% households was
between 25,000 and 49,999 (HK dollars), 19.7% households were earning less than 9,999
(HK dollars) per month, and 10.4% households were in the highest-earning category with
monthly income more than 50,000 (HK dollars).
3. A binary logistic regression model as the inferential analysis was used to estimate
self-reported diabetes on age, sex (referent female), and monthly household income.
Household income, in the regression model, was considered referent to an income group of
above $ 50,000. Hence, in comparison concerning the highest income group was conducted
for analysis of diabetes prevalence in lower-income groups. The age of participants, in the
logistic regression model, was considered in referent to 0–39 years. Here, a comparison
regarding the occurrence of the diseases was achieved for the older people about the youngest
bunch of people.
Binary logistic regression was used since the outcome or dependent variable (self-
reported diabetes occurrence) was a dichotomous variable with male and female as two
categories (Sarkar, & Midi, 2010). The impact of age and household income on self-reported
diabetes was compared between the two genders.
4. A drastic rise in self-reported diabetes was noted for the adjusted prevalence in both
sex groups. Across the years 2001–2008, a gradually increasing self-reported diabetes level
was noted for people aged over or equal to 75 years.
The logistic regression model 2001 was used as the reference point. Controlling for
age, the relative increase rate in diabetes among male adults (P = 27.8%, 2005; P = 47.9%,
7

2008) was lower compared to female adults (P = 31.8%, 2005; P = 69.3%, 2008) (p <
0.001). Diabetes prevalence gradually increased across years in the lowest income groups for
both the genders, after adjusting for age and sex. Hence, low income was noted to be a
significant factor in diabetes occurrence. However, no proof of a definite difference between
prevalence trends was found between male and female respondents.
5. Analysis of the adjusted odd-ratios revealed the association between the adjusted
levels of age, sex, and household income. A strong positive association between the referent
age group (0-39 years) and the other two age groups (40-65, > 65) was noted. Compared to
referent age group, people in age group of 40-65 years (AOR = 32.21, 95% CI 20.6–50.4, p <
0.001), and more than 65 years (AOR = 120.1, 95% CI 76.6–188.3, p < 0.001) were found to
be highly and significantly affected by diabetes. Again, compared to the referent income
group (≥ 50, 000 HK dollars), the lowest income group (AOR = 2.19, 95% CI 1.66–2.88, p <
0.001) was significantly related with diabetes prevalance. Gender of the respondents’ had no
association with the prevalence of diabetes. Therefore, the lowest income group of people
had more than twofold chances of associating with diabetes prevalence compared to people
from the highest income level.
6. The logistic regression analysis was able to explain a mere 20% of diabetes
prevalence. The research required more variables to investigate the relationship with risk
factors of diabetes in the large population of China. The study was based solely on self-
reported data to determine the prevalence of diabetes. Interestingly, some of the previous
studies found that not all people measure diabetes regularly. Therefore, an anomaly in
diabetes prevalence would have been a concern regarding the validity of the readings.
Research by Gu et. al. (2003) suggested that almost 75% of diabetes patients in China were
8

undiagnosed, confirming the reason behind possible anomaly in diabetes prevalence records.
Moreover, external factors such as BMI, lifestyle, family history of the patient regarding
diabetes were not considered in the present study. Again, a stratified random sampling
methodology can generate more precision compared to random sampling. Hence, the results
obtained from the study can be considered trustworthy if self-reported data can be cross-
verified.
References
Gu, D., Reynolds, K., Duan, X., Xin, X., Chen, J., Wu, X., ... & Inter ASIA Collaborative
Group. (2003). Prevalence of diabetes and impaired fasting glucose in the Chinese
adult population: International Collaborative Study of Cardiovascular Disease in Asia
(InterASIA). Diabetologia, 46(9), 1190-1198.
Lam, T., Liang, W., Chikritzhs, T., & Allsop, S. (2014). Alcohol and other drug use at school
leavers' celebrations. Journal of Public Health, 36(3), 408-416.
Mood, C. (2010). Logistic regression: Why we cannot do what we think we can do, and what
we can do about it. European sociological review, 26(1), 67-82.
Pettigrew, S., Biagioni, N., Daube, M., Jones, S. C., Chikritzhs, T., & Kirby, G. (2015).
Understanding and addressing the Schoolies phenomenon. Australian and New
Zealand Journal of Public Health, 39(2), 135-140.
Sarkar, S. K., & Midi, H. (2010). Importance of assessing the model adequacy of binary
logistic regression. Journal of Applied Sciences, 10(6), 479-486.
9