1
PSYC30013 Research Methods for Human Inquiry
Exam: Semester 1, 2022

Available: 14 June 2022 at 8am AEST
Due: 17 June 2022 at 11:59pm. Absolutely no late exams accepted so submit as you go!
Questions: 30
Word count: 2000
Total marks: 65
Pages: 14

Hello everyone! This is the description for the exam, which is due on Canvas on Friday June 17th,
2022 at 11:59pm. You’ll need to submit a Word-knitted version of the completed R Markdown file
found in this zip file, according to the following instructions:

1. Rename the document called exam.Rmd as studentID-exam.Rmd. (Replace studentID with your
student ID number). This is your R Markdown file, where you’ll be putting all your code and answers.
Replace “Your name, Student ID” in the header of the R Markdown file with your name and student
ID. (Keep the quotes)

2. While we encourage collaboration in tutorials and learning in general, you should not be
collaborating with anybody AT ALL for this assignment. That means sharing code privately or
publicly; even talking in the abstract about problems will effectively be collusion. You
should be completing it independently, with no help from any other person in any capacity. Of
course, as always, you are free to use any of the resources from the class to help you, and you're also
free to google or look anything up that you like (as long as you aren't asking anybody, including
anybody on discussion boards, specific questions about this assignment, and everything you write is
in your own words). Different people will be getting different datasets and thus may have different
answers even on the same questions anyway.

3. Plagiarism check is enabled and you can check the similarity report on your submission. In
previous years we have found people who tried to cheat, so please don’t risk it! That said, understand
that we will not be naively looking at the overall % figure: with this sort of assignment a certain
amount overlap is inevitable, so don’t worry if you get what looks like a high % score. Probably most
people will. We will be using the plagiarism check for the parts of the assignment where we'd expect
some variability, and to give a general sense of the overall gestalt.

4. Complete all of the problems below in the R Markdown document. Do not change any of the
arguments to the code chunks, like the names of the code chunks or where it says echo=FALSE or
whatever. If a problem asks you to display a tibble or variable so it shows up in the knitted version,
make sure that you do as the marker cannot evaluate it without seeing it, and if they can't see it
then they won’t be able to award you points for it! Remember that to display a tibble (or any variable)
you just type its name on a line of its own within the R chunk.

5. I've structured this so that, as much as possible, questions do not build on each other. That means
that if, say, you can't get Q5 then you can still get Q6. Try to do all of them.

6. Go for partial credit! Many of these questions have some form of partial credit possible. What that
means is that if it is asking for some R code, break down the problem into pieces. Even if you can only
do some of the pieces, or do them part of the way, that will be worth something. [Note that there is no
question-by-question rubric available because designing one would mean giving away the answers. In
general we will give full credit for responses that correctly address all of the parts of the question.]
Short answer questions (SAQs) are also worth partial credit and are generally asking for some
thoughtful interpretation. If it is based on a previous graph or test you've done, if you did the first
part wrong but discuss it well, you can still get most or all points for the SAQ part. If your code does
not run but you want to include it for possible partial credit, just comment it out (using the # sign) so
that it shows up in the knitted document. If you include a lot of commented-out code and some is
correct and some isn’t, we will not give you credit for the commented-out code; put the thing in there
that you think is the closest to the correct answer, don’t just include everything.
2

7. We are not overly worried about to what decimal place you round answers to and unless the
problem explicitly specifies what you should round to, you will not lose credit for this unless you
round so much that your answer is impossible to discern (e.g., don’t round p-values to the nearest
integer!). Similarly, you will not lose points for trivial presentation things like using parentheses
instead of commas around statistical references, as long it’s clear. That said, for those who want a
guideline, I’ll suggest that you follow APA format or round p-values to three decimal places, degrees
of freedom to one, and test statistics and probabilities to two.

8. If the question is an SAQ, it specifies a word count. Unfortunately, there is no way that I know of in
R Markdown to do a word count. You need to either calculate it from the knitted document, or type
up your answer in Word1 and then cut and paste it into the R Markdown file. (Please put your answer
in between the word ANSWER and [Word count: XX]; needless to say, those two bits do not count
towards your word count.) The word counts I've specified in each question are designed to give you a
guideline about the number of words you should need answer completely and correctly. So don’t feel
like you must use all of the words; if you can answer it fully with less, that’s fine. It’s okay to go
over the word limit for individual questions as long as the total word count for all of the
questions is under 2200 words (i.e., 2000+10%, with the standard penalty if it is 2000 + 10% or
higher2; there is no penalty for having fewer words than 2000-10%). Also, if a question hasn’t
specified a word count then you don’t need to include it in your total word count; these are questions
that require very few words so it’s not worth the effort.

9. There is no word count for code chunks. Word count only applies to the SAQs. Remember to
report your total word count for the assignment as a whole at the top of the document. Your total
word count is the sum of the word counts for all of the SAQs.

10. You'll be turning in the knitted output of your R Markdown file. We prefer that you knit to Word
but if you can't get Word to knit then html is okay. In the worst case, you can turn in the completed
Rmd file. We highly, highly recommend that you knit as you go: (a) knitting can identify
problems in your code that you would have otherwise missed; and (b) you do not want to get close to
the deadline and think you’re done only to find that you’re having troubles knitting. Save yourself the
panic and knit often.

11. Similarly, you can turn in the exam multiple times, so please turn it in even before it’s perfectly
polished. That will save you last-minute panic or computer issues. Also, take a screenshot for proof of
having turned it in just in case you need it. If you do not submit your exam by the due date we
cannot accept it no matter what so just submit as you go! We will only mark the most recent
submission so you lose nothing by doing so.



The story

For the exam, we return to Bunnyland and Otherland one year later! When we last saw them things
had come to a tentative conclusion but substantial challenges remained. Could people from both
lands manage to work together to solve their food problem? Would tensions flare? Would the fragile
peace get disrupted? Would they be able to grow enough food that nobody went hungry?

You'll find answers to these questions in the vignettes below. We'll visit each of our friends and see
what they're doing and you'll be able to help them with their statistics one last time. Enjoy!

* * *


1 I know different software calculates word count in slightly different ways, so we are using Word as the standard.
2 The penalty is as detailed in the student manual: “10% of the total marks available for a given assessment task will be
deducted for every 10% that the word count exceeds the word limit specified for the task”. So up to 2,199 results in no
penalty. Between 2,200 and 2,399 is a 10% deduction, between 2,400 and 2,599 is 20%, and so forth.

3

Punk music

The last year has been a good one for Flopsy. They have embraced their non-binary identity and
decided to change their name and pronouns to something that felt more like "them". They are now
known as Flaye and have formed a goth punk band along with Sissily the snake. They've spent much
of the last year touring around Bunnyland and Otherland with their band, which is called Mental
Limit Theorem. They play in four venues: Bunnystadia, Star Venue, Zed Zone, and Markov Chain.

Flaye and Sissily are curious about whether more people tend to attend their concerts at any of the
venues, or whether all four of them have about the same average attendance. Fortunately they have
tracked the concert attendance, which can be found in the tibble called df. It has been loaded for you.
Each row consists of one concert, and there are three columns:

concert: unique identifier for each concert
venue: one of the four possible venues (Bunnystadia, Star Venue, Markov Chain, or Zed Zone)
attendance: the number of people in the audience at that concert

Their research question is whether on average attendance is the same at each of the four venues. Your
job is to use your R and statistics knowledge to find the answer to this question.


Q1 [2% of total marks]

First, as always, let's visualise the data. In the code chunk for this question make the appropriate
figure visualizing the attendance for each of the venues. Your figure should show each of the
individual data points as well as a different geom displaying a summary for each venue, a colour
scheme (other than the default) with different colours for each venue and the colours of the data
points matching the geom, no legend, and of course a title, informative axis labels, and a nice theme.

Q2 [2% of total marks]

There are several assumptions you could test, but for the sake of simplicity we will focus on the
normality assumption (you may presume that homogeneity of variance holds). Evaluate normality
using the appropriate test. In 30 words or less, report the statistical test and indicate what the results
of the code reveal (including stats reference). Was the assumption violated? How can you tell?

Q3 [4% of total marks]

Run the appropriate statistical test to evaluate the research hypothesis given Q2. Report on the
results in 60 words or less. In your report, don't worry about including descriptive statistics but do
include an explanation of which statistical test you used, what the predictor and outcome variables
were, the appropriate stats reference, the effect size and its interpretation, and the interpretation of
this data in terms of the research question.

Q4 [4% of total marks]

Perform post-hoc pairwise tests and in 40 words or less, report the tests and which venues (if any)
had significantly lower attendance than others, along with their p-values. You do not need to report
the non-significant venues or their p-values.







END OF SECTION
4


Art

Not much has changed for Cuddly Paws, although all the drama made her think even more about
what she wanted out of life. She hadn't enjoyed all of the tension and worry but had enjoyed how
everything had forced her out of her shell a bit and deepened her interactions with people. Though
she still preferred the comforts of art over people, she did expand out of her comfort zone enough to
begin giving her art away.

It turns out that many people enjoy her art, and some of them react by sending her letters to say so.
Some of the letters are purely kind and complimentary, while others contain some kind of question or
request, like asking her if she would paint in a certain style or show her art in a gallery. No letters
contain both compliments and requests. Cuddly Paws spent a few weeks keeping track of these letters
and eventually found herself wondering if the type of letter was different depending on who was
sending it (bunnies or bears). The data can be found in the table called artTable, which has been
created for you in the code chunk for Q5. It is also shown below.



Cuddly Paws would like to know if bunnies and bears differ in the distribution of compliments and
requests in their letters, based on this data.

Q5 [3% of total marks]

Run the appropriate statistical test to evaluate Cuddly Paws’ question and report on the results in 80
words or less. In your report, don’t worry about including descriptive statistics, but do include an
explanation of which statistical test you used and why, the appropriate stats reference, and the
interpretation of this data in terms of the research question. Don’t worry about effect size.

























END OF SECTION
5
Meatball Cafe

The opening up of relations between Bunnyland and Otherland has been great for Gladly. He hadn't
realised how hard he found it as one of the few bears in a land full of bunnies until things changed.
He ended up making fast friends with Super Size, who also had struggled with a similar alienation
being so large and different. He was still large and different, but with Gladly that didn't seem to
matter, and the two bonded over one of their favourite topics: food.

Now that they were growing so much more food, both Gladly and Super Size's persistent hunger was
assuaged, and the two friends turned their thoughts to how they could share their love of food with
others. One day Gladly had an idea: they could start a restaurant! It was ideal for both of them. They
could be around friends all day, and make use of all of the fantastic new foods being grown.

Before long, their restaurant was the social hub linking Bunnyland and Otherland. Located on the
border, Meatball Cafe was a vibrant place of food, conversation, and laughter. No longer did Gladly
wonder what he wanted to do with his life: he knew! And best of all, there were meatballs every day.

Before long they started wondering what factors were associated with more sales, so they tracked this
information over the course of the spring. The data can be found in the dm tibble, which has been
loaded. Each row corresponds to the data from one day. There are four columns:

day: What day it was (day1 = September 1, day91 = November 30)
temperature: The maximum temperature that day in degrees C
happiness: The happiness of the person serving that day (0 = very sad; 50=very happy)
meatballs: The number of meatballs sold on that day

Their research question is what factors (if any) predict how many meatballs are sold: the temperature
that day, the happiness of the server, and/or an interaction of the two. Your job in the next few
problems is to use your R and statistics knowledge to find the answer to this question.

Q6 [7% of total marks]

Run the appropriate statistical test to evaluate the research hypothesis, assuming for now that no
assumptions have been violated. Report on the results in 170 words or less. In your report, don't
worry about including descriptive statistics but do include an explanation of which statistical test you
used, the predictor and outcome variables, the stats reference, and the interpretation of this data in
terms of the research question. Include effect size and a description and interpretation of all the
significant unstandardised coefficients except the interaction (whether it is significant or not). Don’t
worry about calculating or reporting standardised coefficients.

Q7 [6% of total marks]

There are four things we discussed in lectures that you should check for the test you just ran. Check
them in the four code chunks (the order does not matter). For each, indicate what checked, what the
results indicate, and how you can tell. Use no more than 180 words total for this entire question.

Q8 [5% of total marks]

Given the results in Q6 and Q7, what would you do to answer this research question? You may use the
code chunk to make any figure(s) you think necessary to figure out what is going on. You do not need
to perform any additional tests or analyses, but you should say what you think is going on as well as
you would do and why if you were in this situation in real life. This could include describing any
additional analyses that you would perform and how they would support different conclusions. Use
no more than 240 words. [Note: this is a challenging question and you are being asked to critically
think about the data and the rest of the process. Marks depend primarily on the clarity,
appropriateness, and thoroughness of your discussion. There are multiple ways to earn full credit.]

END OF SECTION
6

New leaders

The path toward integrating the people of Bunnyland and Otherland was not easy, but was helped a
lot by the burgeoning friendship between LFB and Rainbow the unicorn. LFB will always remember
Rainbow as her first friend in Otherland. For her part, Rainbow has found LFB to be a delightful
companion and loves her fresh views and honest, straightforward manner. The trust between the two
of them carried the day several times when tensions were high.

As a result, when it came time for the next elections, there was a groundswell of support for the idea
of electing LFB and Rainbow together to be leaders of their respective lands. LFB and Rainbow were
more surprised than anyone, but found that they truly enjoyed being able to guide their people and
work together to make things better.

It has now been five months since they were elected and Rainbow and LFB would like to know
whether people still support them as leaders. They therefore commission a poll which surveys 200
different people, 100 from Bunnyland and 100 from Otherland. 50 people from each land were asked
to rate their degree of support for LFB, and 50 were asked the same question about Rainbow; nobody
was asked both questions. This data can be found in the ds tibble, which has been loaded for you.
Each row consists of one person. There are four columns:

id: Random code corresponding to each person who took the survey
location: Whether that person lives in Bunnyland or Otherland
leader: Whether they were asked about LFB or Rainbow
support: Level of support for the leader they were asked about (0=none; 100=full)

LFB and Rainbow want to know whether the two of them have the same level of support, whether
support varies by location, and whether there is any interaction between these factors.


Q9 [3% of total marks]

First, as always, let's visualise the data. In the code chunk for this question make a bar plot showing
the support for each person. You should have two panels, one corresponding to Bunnyland and one
for Otherland. Within each panel, the two bars should be semi-transparent and of different colours
(not the default) and should how the mean support for LFB and Rainbow respectively (with error
bars corresponding to standard error) and individual data points in the same colour as the bars. Your
figure should also have no legend, a title, informative axis labels, and a nice theme.

Q10 [6% of total marks]

For simplicity let's assume that all the assumptions of the relevant statistical test have been met, so
you don't have to evaluate them. Instead, run the appropriate statistical test to evaluate LFB and
Rainbow’s research question and report on the results in 120 words or less. In your report, don't
worry about descriptive statistics but do include an explanation of which statistical test you used, the
appropriate stats reference, and the interpretation of this data in terms of the research question.
Don’t worry about effect size or any post-hoc tests.


END OF SECTION








7

Psychology

Following the events of a year ago, Bunny achieved her dream and became a psychologist. She also
grew to be good friends with Foxy, having been impressed by the empathy and courage that became
visible during her adventures in Otherland. That adventure and the bond with Bunny gradually led
Foxy to come out of her shell more, sharing more about her upbringing in far-away Foxland. Upon
graduating with her degree, Bunny and Foxy decided to open up a psychological practice, with Bunny
as the psychologist and Foxy as the manager. Soon their practice -- called Hearts and Minds --
became central to helping the people of Bunnyland and Otherland begin to process the years of slow
trauma caused by increasing hunger, fear, and uncertainty. And both Bunny and Foxy grew greatly in
confidence, having faced so many of their biggest fears (including statistics and R).

Foxy is still worried a lot, though, and most recently she has started to worry that she is not as
efficient or good as Bunny at being a therapist. She therefore decides to use her skills to help answer
the question. She is interested in how many minutes total each patient took before they felt they were
cured and no longer needed to be seeing a psychologist.

The dataset dp, which has been loaded for you already, contains a list of the last 40 of each of the
patients each of them have had who finished all of their sessions. Each row is one of the patients, and
there are the following three columns:

patient: unique identifier for each patient
person: the person acting as psychologist for that patient (Bunny or Foxy)
time: the total time in minutes they spent in sessions before leaving

Foxy wonders if she and Bunny are significantly different in terms of how long their patients take to
feel cured, on average. She hopes to use this knowledge to infer whether she is as good as Bunny.


Q11 [1% of total marks]

Check the assumption of normality in the code chunk below. In 40 words or less, report the statistical
test and indicate what the results of the code reveal (including stats reference(s)). Was the
assumption violated? How can you tell?


Q12 [7% of total marks]

Run the appropriate statistical test to evaluate the research question and report on the results in 170
words or less. In your report, include descriptive statistics (means, not sd ) and effect size (including
interpretation) as well as an explanation of which statistical test you used and why, the appropriate
stats reference, and the interpretation of this data in terms of the research question. What does this
result indicate, if anything, about whether Foxy is as good of a psychologist as Bunny, and why?

END OF SECTION













8
Adventurers

His missions to Otherland sparked in Doggie a taste for adventure: while all the other people started
settling in and working together, he grew increasingly restless. Moreover, he realised that some early
life trauma meant he would probably not be able to trust the Others for a while, so he decided it
would be best to heal on his own. He spent more and more time wandering around in the hills
nearby, wondering what he wanted to do with his life. On one of these trips he came across Kevin and
Kevin Clark (the string), who were feeling similarly at sea. They got to talking and realised that what
they all wanted was to strike out somewhere new -- to see what there was to see in the mountains and
valleys beyond both Otherland and Bunnyland. They thus decided to give it a try together. Before long
the three of them gained a reputation far and wide as almost mythical adventuring figures.

One of their favourite things about their new life is that they sleep far better than they ever have
before. Perhaps this is because of so much travelling, perhaps because they are eating better than they
used to, perhaps both, perhaps neither. In order to investigate this they decide to track all of these
variables to see if there are any relationships between them. The data is in the tibble dd, which has
been loaded for you. Each row corresponds to one night and it has three columns:

night: the night in question
sleep: the amount of time Doggie spent sleeping in hours
distance: the distance in km Doggie travelled the previous day
food: the number of kilojoules Doggie ate the previous day

Q13 [4% of total marks]

Calculate a Pearson correlation between distance and food and report on the results in 60 words or
less. In your report, use the appropriate stats reference and interpret what this result suggests about
the relationship between distance and food. How does the correlation you found relate to the R2 you
would obtain if you ran a regression with distance as the outcome and food as the predictor?

Q14 [5% of total marks]

Their main question is about how these variables are related to sleep and specifically which of them
matter, if any. For this we are going to do model selection. For simplicity let's assume that all
assumptions are met. In the code chunk below, create the following three models (you don't have to
show the output, just create them).

1. modelSD: sleep is the outcome variable, distance is the only predictor
2. modelSF: sleep is the outcome variable, food is the only predictor
3. modelSDF: sleep is the outcome variable, distance and food are both predictors,
but there is no interaction

Q14a. Perform model selection between these three using AIC or BIC (either is fine) as the
complexity penalty. In no more than 15 words, state which model is preferred and why (you don’t
need to report the AIC/BIC results, just interpret them).
Q14b. In no more than 85 words, explain why we use something like AIC or BIC to do model
selection, rather than just picking the model that has the highest R2.

Q15 [6% of total marks]

In 130 words or less, report the best-fitting model from Q14, including a description of which
statistical test you used, your predictor and outcome variables, the appropriate stats references, and
the interpretation of the research question in terms of this data. Don’t report effect size or the
interpretation of the unstandardised coefficients but do discuss the relative strength of the predictors
and why the size of the values of the undstandardised and standardized coefficients are different.


END OF SECTION


9
Statistical consulting

Finally, we get to Shadow and Little Blue. This semester showed both of them how much they love
statistics and R, and it showed everyone in Bunnyland and Otherland how valuable statistics and R
are when making decisions. As a result, they decided to open a statistical consulting business. Their
job is both to collect data but also to answer anybody and everybody's statistical questions. They are
very happy.

What follows are a few of the questions they've gotten recently. Shadow and Little Blue have gone on
leave, so see if you can answer them as well!

Q16 [5% of total marks]

Q16a. Which of the following statistical references is impossible? For (i) and (ii) indicate your
answer by reporting A or B. Use your knowledge about each test statistic and how it is related to
degrees of freedom and p-values to answer this question. You do not need to do any coding.
(i) A. χ2(3) = 75.3, = .936 or B. χ2(1) = 1.44, = .231
(ii) A. t(44) = 0.67, p = .003 or B. t(35) = -2.68, p = .011

Q16b. For each answer in Q16a, explain it by describing intuitively what the test statistic
captures and thus why you chose the answer you did, making reference to the degrees of freedom
and/or p-value as appropriate. Use no more than 130 words in total for this question.


Q17 [4% of total marks]

Q17a. For each of the following equations,
indicate which figure on the right it corresponds
to. (Note: you do not need to code anything for
this problem; just use your understanding about
what the various coefficients mean, and pay
special attention to the positive and negative
signs).
(i) = 1.51 +
(ii) = 1 + 2 + 12 +
(iii) = 21 − 1.52 +
(iv) = 21 + 2 − 312 +


Q17b. For each answer in Q17a, explain why you
chose it by discussing the relevant aspects of the
equation and how they map onto the figure. You
do not need to discuss every single part of the equation and the figure, only the ones that were
relevant for distinguishing them. [Note: we are not looking for answers of the form “I ruled out all of
the other possibilities, so it must be this one”; although you can mention this if it was part of your
reasoning process, each answer must make reference to that equation and the figure in some way].
Use no more than 150 words in total for this question.

Q18 [2% of total marks]

Quackers has been struggling with statistics – he thinks it is annoying for some reason! – so he
decides to ask you one of his questions. “I don’t understand something about non-parametric tests
like the Wilcoxon,” he says. “I get that we should use them if assumptions of normality are violated
and we can’t use things like the Welch t-test. But why not use them all of the time? If they make fewer
assumptions, aren’t they better?” Quackers goes on, “I asked Shadow and Little Blue about this but
they were in a rush and just muttered something about Type 2 error and power, which I didn’t catch.
So can you explain it to me, including what power and Type 2 error has to do with it?” You have 110
words or less.

10
Q19 [3% of total marks]

Consider all of the figures below. For each of them, explain whether that figure is or is not consistent
with obtaining an F statistic less than one on that data. Your answer for each should be either “is
consistent” or “is not consistent” and thus does not contribute to the overall word count.
























END OF SECTION




























11

Psychological Assessment


Some of these questions require numerical answers only. Don’t delete anything from the Markdown
file aside from the "_______" characters which appears in some places, and which indicates a gap
which you should fill.

Before attempting these questions please ensure that you have studied:
- The Week 11 and 12 lectures
- The Week 12 tutorial
- The Practice Quiz entitled "Revision Quiz for Section 12.3 of Week 12 (Day 2)"
- The questions and solutions contained within the document "pa-exam-prep-prompts-
solutions.pdf", which is linked from the Exam page of the Canvas website for this subject.


Q20 [2% of total marks]

A treatment for Lazar's Disease is being developed. People participating in a study were randomly
selected to receive either the treatment, or else a placebo. Later on, a test was administered to
investigate whether they got the disease.
550 people received the treatment and got the disease
450 people received the treatment and didn't get the disease
430 people received the placebo and got the disease
570 people received the placebo and didn't get the disease

Produce R code to calculate the Risk Ratio and fill the gaps in the Markdown document. Hint: One
gap should be replaced with the word "Placebo", one with the word "Treatment", and one with the
Risk Ratio. Report the Risk Ratio as a proportion, rounded to three decimal places (e.g. 0.991, or
3.120).


Q21 [2% of total marks]

Q21a. Which of the following would most increase your suspicion that you are encountering an
instance of the file drawer problem?
A. The small studies in your funnel plot have effects that vary to an implausibly low degree,
and are spread roughly evenly on each side of the overall average
B. The small studies in your funnel plot have effects that vary to an implausibly low degree,
and are not spread roughly evenly on each side of the overall average
C. The small studies in your funnel plot have effects that vary to an implausibly high degree,
and are spread roughly evenly on each side of the overall average
D. The small studies in your funnel plot have effects that vary to an implausibly high degree,
and are not spread roughly evenly on each side of the overall average

Report A, B, C, or D (and nothing else) in the blank provided on the Markdown file.

Q21b. Let's say you believe you have encountered the file drawer problem. Briefly describe one action
you could take to try to ensure that the file drawer problem does not invalidate your meta-analysis.
More than one reasonable approach was discussed in the lecture, but please describe one action only.
Use 20 words or fewer.







12


Q22 [2% of total marks]

In the Robo-Hungarian Empire the ability to bend metal objects is regarded as an important skill.
There is a test of metal-bending ability which has a reliability of .95, a mean of 20, and a standard
deviation of 3. Flavia scored 22 on this test. You are a government official within the Robo-Hungarian
Empire, and are responsible for making a decision on whether Flavia has showed sufficient skill to
gain employment within your department. You are wondering whether we have good grounds for
believing that Flavia's metal-bending abilities are above average.

Report a 95% Confidence Interval for Flavia’s true score. Give two numerical values: the lower and
upper bounds of the Confidence Interval. Show your workings in the code chunk. Report each value
to 3 decimal places, e.g. 99.123.


Q23 [1% of total marks]

Your workplace is planning to make important decisions about individuals on the basis of their scores
on a test which it has developed. The test consists of 80 items, and its reliability is .88.

Your boss knows that this level of reliability does not meet the desirable standard for making
important decisions about individuals with respect to specific test scores, and is open to increasing
the number of items on the test so as to meet that desirable standard.

Your boss wants you to provide an estimate of the total number of items the revised test will need to
have in order to surpass the desirable standard. Your boss wants the revised version of the test to
have as few items as possible, while nonetheless surpassing the desirable standard.

Provide R code to calculate the total number of items that will be required on the revised version of
the test. Please note that since fractional items are not possible, you should round your answer in an
appropriate manner so as to satisfy the requirements of the question. You should either report your
answer as an integer, or with only zeros after the decimal place.

Hint: Your answer must necessarily be higher than 80, since the original test already has 80 items.


Q24 [2% of total marks]

You understand that increasing the number of items is not the only way to improve the reliability of a
test. Suggest two other ideas which would have the effect of increasing the reliability of the test.
Please label your ideas "Idea 1:" and "Idea 2:". Suggest only ideas that were addressed in the lecture,
and don't suggest increasing the number of items, as we already covered that idea in Q23. Respond in
50 words or fewer.


Q25 [4% of total marks]

There is a disorder named Glossop’s Disorder. Individuals over 40 years of age often develop serious
symptoms as a result of Glossop's Disorder. We will not discuss these symptoms in full here, but they
include things such as developing a tendency to put random words in “quotation” marks when
writing, as well as a tendency to inappropriately end sentences with ellipses...

In young people between the ages of 18 and 25 the only known symptom of Glossop's Disorder is that
individuals with the disorder tend to sleep for 9 to 12 hours per night on average. The majority of
young people who average 9 to 12 hours of sleep per night do not have Glossop's Disorder. However,
the majority of young people with Glossop's Disorder average 9 to 12 hours of sleep per night.

13
A test for Glossop’s Disorder has been developed, and a trial was done in which the test was applied
to 100,000 randomly selected 21-year-old members of the population. The trial results showed:
95 true positives (people who were criterion positive, test positive)
220 false positives (people who were criterion negative, test positive)
99680 true negatives (people who were criterion negative, test negative)
5 false negatives (people who were criterion positive, test negative)

Give a numerical answer to each of questions in part (a)-(d). Express all these numerical
values as percentages, with each percentage reported to three decimal places (e.g.
99.123%). Use the code chunks to show the calculations for each.

Q25a. What is the test sensitivity?
Q25b. What is the test specificity?
Q25c. What is the test Positive Predictive Power?
Q25d. What is the test Negative Predictive Power?


Q26 [2% of total marks]

This question relates to the situation and data described in Q25.

You are a specialist in sleep disorders, and in that context a 21-year-old named Alexei has been
referred to you. Alexei is concerned because over a prolonged period he has been sleeping 11 hours
per night on average.

You administer the test for Glossop's Disorder to Alexei, and he tests positive.

In 30 words or fewer, do you estimate that the probability Alexei has Glossop’s Disorder is
- higher than the test Positive Predictive Power stated in your answer to Q25c
- lower than the test Positive Predictive Power stated in your answer to Q25c
- the same as the test Positive Predictive Power stated in your answer to Q25c

In your answer you must re-state one of these three options and also briefly explain the reasoning
behind your answer.


Q27 [1% of total marks]

You are a practitioner and are choosing between the following two tests

Test A: 51.000% sensitivity, 97.001% specificity
Test B: 96.000% sensitivity, 97.000% specificity

In the context in which you find yourself, the key goal of testing is to "rule in" the presence of the
disease, by which we mean to obtain strong evidence that the disease is present. The reason why it's
important to "rule in" the presence of the disease is that persons who test positive will be selected for
additional procedures. These additional procedures are quite invasive and expensive, and it's
important that they are not unnecessarily applied to people who are free of the disease.

With that goal in mind, which test would be the better one for you to choose, Test A or Test B? For the
purposes of this question you won't have any access to any other information about the test, even
though such information would have been potentially helpful - thus, for example, you won't have data
on PPP, NPP, or prevalence. For this question, the only possible answers are "We should prefer Test
A" or "We should prefer Test B", and you should not explain the reasons for your preference.




14


Q28 [1% of total marks]

Imagine that you are taking a university subject with a stated aim of covering French history during
the period 1800-1919, with one decade covered in each week of the course. The only assessment for
the subject is an exam in which you are required to write an essay on one decade of your own
choosing (so, for example, you might choose to write about French history during the years 1880-
1889). What is the term for the threat to validity that is posed by the scenario described above?
Answer in 10 words or fewer.


Q29 [3% of total marks]

Within the context of a Multitrait-Multimethod Matrix, consider the following two correlations:

Correlation A. Correlation between self-report measure of impulsivity and interviewer rating
measure of impulsivity.
Correlation B. Correlation between interviewer rating measure of social skill and
acquaintance report measure of impulsivity.

Q29a. Would you prefer that Correlation A is higher than Correlation B, or that Correlation B
is higher than Correlation A, or that the two correlations are about the same? The only permissible
answers here are "I would prefer Correlation A be higher than Correlation B", "I would prefer
Correlation B be higher than Correlation A", or "I would prefer that both correlations are about the
same". Therefore you must answer in exactly 10 words.
Q29b. What sources of variance does each correlation (i.e. Correlation A and Correlation B)
relate to? Answer in 70 words or fewer.


Q30 [1% of total marks]

This is a freebie - as long as you say anything, you will get full credit! Who was your favourite
character in Bunnyland or Otherland, and why?



END OF EXAM