GEA1000 QUANTITATIVE REASONING WITH DATA
TUTORIAL 2
Please work on the problems before coming to class. In class, you will engage in group work.
Case Study 1: Working with categorical variables
The data set [Link] comes from a study conducted on penguins in the Antarctic
region. (Recall that you have encountered this data set in a technical video from the previous
chapter.) Suppose you are a researcher working with a marine biologist. For this case study,
let our guiding research question be: For the Biscoe region, which species of penguin is
associated with being underweight? (Assume that the minimally healthy weight is 3.62kg1 for
the Adelie species and 4.50kg2 for the Gentoo species.)
1. Filter the data set to only display entries from the Biscoe region.
a. Which species of penguin are we comparing in this case? Give a breakdown of
their proportions. (Express them as percentages correct to 1 decimal place.)
b. Which species of penguin is associated with being underweight? (Hint: You
may consider making use of the IF function in Excel.)
c. After finding the association between species and being underweight, the
marine biologist asks: Is sex a confounder in this case? (Note: You will find
instances of “NA” entries for Sex in the data set – remove those data points as
part of your data cleaning in this case.)
1
Source: [Link]
2
Source: [Link]
Case Study 2: Confounders and Simpson’s Paradox
The file [Link] contains information on the enrolment of university students in
Singapore in 2002 and 2022, with the variable First Degree describing the first degree course
a student is enrolled in. Use appropriate software to answer the following questions, giving
your answers to 2 decimal places.
2. Let the degrees ‘Engineering Sciences’ and ‘Natural and Mathematical Sciences’ be
regarded as STEM degree types, and the other degrees be regarded as Non-STEM
degree types.
a. Calculate the proportion of students enrolled in STEM first degrees in the data
set.
b. Is there any association between year of enrolment and whether a student is
enrolled in a STEM first degree? What about that between sex and whether a
student is enrolled in a STEM first degree?
c. Is sex a confounder when studying the association between a student’s year
of enrolment and whether a student is enrolled in STEM first degree?
d. Upon completing parts (b) and (c) of this question, your friend Tammy told you
that the answer to part (c) could have been deduced from (b). She explains
that sex must be a confounder in the study because it is associated with
whether a student is enrolled in a STEM first degree, which is then associated
with the year of enrolment. Hence, sex is associated with both a student’s year
of enrolment and whether a student is enrolled in a STEM first degree. Based
on Tammy’s logic alone, is Tammy’s claim necessarily correct?
e. In relation to part (c), do we observe Simpson’s Paradox when examining the
association between a student’s year of enrolment and whether a student is
enrolled in a STEM first degree?
Case Study 3: Confounders in an experimental study
Background: Polio, also known as infantile paralysis, is an infectious disease that strikes
young children, often causing permanent paralysis. It spreads through person-to-person
contact. In the 1950's, American scientist Jonas Salk developed a vaccine that protected
monkeys from polio and was safe when injected into human subjects in the laboratory in the
1950's. By 1954, the vaccine was ready to be tested in the real world.
In order to determine if the polio vaccine reduces the risk of polio infection, a cohort of
children were invited to take part in a study, known as the NFIP study. However, only some
children had parental consent to receive the vaccine, which posed a problem for researchers.
Consider the following two study designs:
NFIP study:
• Children with parental consent: Assigned into vaccinated and control groups
• Children without parental consent: Assigned entirely to the control group (since they
did not consent to taking the vaccine)
Sample size Polio No Polio
Vaccinated 225000 56 224944
Control 725000 391 724609
Exclusion study:
• Children with parental consent: Randomly assigned into vaccinated and control
groups. Children and doctors were blinded via the use of a placebo.
• Children without parental consent: Excluded from study
Sample size Polio No Polio
Vaccinated 200000 56 199944
Control 200000 142 199858
Information on study participants:
• Families who provided consent tended to be of a higher income group and as a
result lived in more hygienic conditions.
• Children living in more hygienic conditions were more susceptible to polio, as they
were not exposed to the virus since young and lacked immunity to the virus.
Use the information above to answer the following questions, giving your answers to 4
decimal places.
3ai) Calculate the conditional rate of getting polio given that they are vaccinated for each
study. 56/225000 56/200000
(ii) Calculate the conditional rate of getting polio given that they are in control group for
each study.
(iii) Comment on the appropriateness of using rates rather than absolute number of polio
cases to compare the effectiveness of the vaccine in the NFIP study. sample size different, if use absolute number, it
will look like a very big number
3b) Evaluate the two study designs by answering the questions below.
(i) To what extent were the study participants randomly assigned into treatment and control
groups?
NFIP study Exclusion study
Random Assignment
not randomly assigned - Children with parental consent were
children in treatment group also not randomly assigned randomly assigned.
- Children without parental consent were
excluded.
(ii) Discuss how this assignment might affect the way we interpret results from both studies.
NFIP study Exclusion study
- As both groups are big, random
- children in control group (never give consent) assume lower
income, assignment makes the features of both
living in not so hygienic condition, exposed to different viruses, more
immunity when fighting polio
treatment and control groups similar.
- The only difference is thus due to the
for the nfip, the control group rate of catching polio is lower than presence/absence of the vaccine.
what it should be (as they are a group with high immunity - This, together with double blinding done,
already),, biased against the vaccine
allows us to conclude that the difference
in rates is most likely due to the vaccine.
(iii) According to the table of results, by how much does the vaccine reduce the polio rate?
Do you think the vaccine is actually more or less effective than what you calculated?
NFIP study Exclusion study
- The vaccine reduces the polio rate by
- the actual effect of the vaccine is likely to be larger than 0.0710% - 0.0280% = 0.0430%
0.029%
- - The actual effect is likely to be similar to
0.043%, since the study was conducted
with randomised assignment, double
blinding and a large sample size.
generalisability problematic as there is exclusion of those
parents who do not consent