lOMoARcPSD|4683239

Lab Test 2018 Answers

Experimental Design And Data Analysis (University of Melbourne)

StuDocu is not sponsored or endorsed by any college or university

Downloaded by Zihan Yan ([email protected])
 lOMoARcPSD|4683239

MAST10011 Computer Test — Semester 1, 2018

Name:

Student Number:

Tutor:

Instructions:
❼ This test contains 2 problems, worth a total of 25 marks.

❼ It accounts for 10% of the final assessment for this subject.

❼ This test will be conducted under examination conditions: you are not allowed to talk
until after you leave the room. You must work individually.
❼ This is an open book test: you may use any printed or hand-written materials. Electronic
devices (including, but not limited to, calculators and mobile phones) are not permitted
and will be confiscated.

❼ Write your answers in the spaces provided. You will need to use R to produce graphs
and statistics in order to answer the questions on this test. You DO NOT need to write
down your commands, or copy any graphs which you produce.
❼ You may only use R and RStudio. Internet access and all other software, including Excel,
are NOT PERMITTED.
❼ You are allowed to use your lecture notes, annotated computer lab sheets and any other
printed or written notes.
❼ You have 40 minutes to complete the test.

DO NOT TURN THIS PAGE until

you are instructed to do so.

Test Number 1

Downloaded by Zihan Yan ([email protected])

 lOMoARcPSD|4683239

Question 1 [12 marks]

Otitis media, a disease of the middle ear, is one of the most common reasons for visiting a doctor in
the first 2 years of life (other than a routine visit). Let X be the random variable that represents the
number of episodes of otitis media in the first 2 years of a baby’s life. In Australia, X has the following
probability mass function and cumulative distribution function:
x 0 1 2 3 4
Pr(X = x) 0.13 0.41 0.27 0.15 0.04
Pr(X ≤ x) 0.13 0.54 0.81 0.96 1

a. [2 marks] Calculate the median and interquartile range of X.

Solution: median = 1; IQR = 1. 1 mark each.
b. [2 marks] Produce a bargraph of the distribution of X on-screen, and comment on its shape.
Justify your comment with supporting statistics.

> p <- c(0.13,0.41,0.27,0.15,0.04)

> barplot(p)
0.4
0.3
0.2
0.1
0.0

Solution: X is clearly right-skewed. This can be seen by the fact that its mean (1.56, below) is
larger than its median (1). 1 mark for right-skewed, 1 mark for justification.

Downloaded by Zihan Yan ([email protected])

 lOMoARcPSD|4683239

c. [2 marks] In a family with 4 children, what is the probability that exactly 1 will have more than
1 episode of otitis media in the first 2 years of their life?

> dbinom(1, 4, 1-0.54)

[1] 0.2897338

Solution: 0.290. 1 mark for binomial, 1 mark for correct answer.

d. [3 marks] From the distribution above, it can be calculated that E(X) = 1.56 and sd(X) = 1.023.
In a town of 5000 babies, calculate a 95% probability interval for the total number of episodes of
otitis media for all of these babies.

> qnorm(c(0.025,0.975),5000*1.56,1.023*sqrt(5000))

[1] 7658.222 7941.778

Solution: (7658, 7942). 1 mark for normal, 1 mark for correct mean/sd, 1 mark for correct
answer.

e. [3 marks] In a study of 200 babies in New Zealand, it was found that 79 of them had more than 1
episode of otitis media in the first 2 years of their life. Test the hypothesis that there is a different
proportion of such babies in New Zealand than in Australia, at the 5% significance level. Clearly
state your hypotheses, p-value and conclusion in the context of the problem.

> prop.test(79, 200, 1-0.54)

1-sample proportions test with continuity correction

data: 79 out of 200, null probability 1 - 0.54

X-squared = 3.1451, df = 1, p-value = 0.07615
alternative hypothesis: true p is not equal to 0.46
95 percent confidence interval:
0.3274613 0.4666408
sample estimates:
p
0.395

Solution: We test H0 : pNZ = 0.46 against H1 : pNZ 6= 0.46. With a p-value of 0.076, we do not
reject H0 : there is insufficient evidence to show that New Zealand has a different proportion of
these babies. 1 mark for hypotheses, 1 mark for p-value, 1 mark for conclusion in context.

Downloaded by Zihan Yan ([email protected])

 lOMoARcPSD|4683239

Question 2 [13 marks]

A new drug, RedChol, is designed to reduce cholesterol levels. A pilot study is conducted of 20 overweight
middle-aged males (45–64 years old, with BMI > 30) who volunteered for the study. These individuals
were randomly assigned to two treatment protocols which were followed for three months:
❼ a low-cholesterol diet, a basic exercise routine and RedChol (12 subjects);
❼ a low-cholesterol diet, a basic exercise routine and a placebo (8 subjects).
The cholesterol levels of the participants (in mg/100mL) were measured before and after the study, and
the difference (after−before) recorded in the table below.
RedChol 6.9 6.6 -10.6 -16.8 3.5 -10.4 -22.1 0.4 -17.2 -21.9 -11.1 -8.5
Placebo -26.8 -15.5 -13.3 -6.2 8.4 1.1 -1.9 3.6
a. [2 marks] Calculate a 95% confidence interval for the average reduction in cholesterol for the
RedChol group.
> redchol <- c(6.9,6.6,-10.6,-16.8,3.5,-10.4,-22.1,0.4,-17.2,-21.9,-11.1,-8.5)
> placebo <- c(-26.8,-15.5,-13.3,-6.2,8.4,1.1,-1.9,3.6)
> t.test(redchol)
One Sample t-test

data: redchol
t = -2.7829, df = 11, p-value = 0.01781
alternative hypothesis: true mean is not equal to 0
95 percent confidence interval:
-15.103220 -1.763446
sample estimates:
mean of x
-8.433333
Solution: (-15.10, -1.76). 1 mark for t-test, 1 mark for answer.
b. [2 marks] What distributional assumption are you making to answer part (a)? Produce an ap-
propriate plot on-screen to check this assumption, and comment on your findings.
> qqnorm(redchol)
> qqline(redchol)

Normal Q−Q Plot

● ●
5

●
0
Sample Quantiles

−5

●
−10

● ●
●
−15

●
●
−20

● ●

−1.5 −1.0 −0.5 0.0 0.5 1.0 1.5

Theoretical Quantiles

Solution: We assume that the difference in cholesterol levels of the RedChol group is normal.
From the QQ plot, this does not seem unreasonable. 1 mark for normal assumption, 1 mark for
QQ plot.

Downloaded by Zihan Yan ([email protected])

 lOMoARcPSD|4683239

c. [3 marks] Calculate a 95% prediction interval for the reduction in cholesterol for someone taking
the RedChol treatment.

> mean(redchol) + c(-1,1)*qt(0.975,df=11)*sd(redchol)*sqrt(1+1/12)

[1] -32.48195 15.61529

Solution: (-32.48, 15.62). 1 mark for mean/sd, 1 mark for formula, 1 mark for answer.

d. [3 marks] Test the hypothesis that RedChol is as effective as a placebo for reducing cholesterol,
at the 5% significance level. Clearly state your hypotheses, p-value and conclusion in the context
of the problem.

> t.test(redchol, placebo, var.equal=TRUE)

Two Sample t-test

data: redchol and placebo

t = -0.42222, df = 18, p-value = 0.6779
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
-12.59913 8.38246
sample estimates:
mean of x mean of y
-8.433333 -6.325000

Solution: We test H0 : µRedChol = µplacebo against H1 : µRedChol 6= µplacebo . With a p-value of

0.678, we do not reject H0 : there is insufficient evidence to conclude that RedChol performs better
than a placebo. 1 mark for hypotheses, 1 mark for p-value, 1 mark for conclusion in context.

e. [3 marks] The makers of RedChol claim that it reduces cholesterol levels by an average of
12mg/100mL in three months. The makers of a rival drug, BlueChol, claim that RedChol reduces
cholesterol levels by only 8mg/100mL. Calculate the sample size needed to distinguish between
these claims with power 0.9 and significance 0.05.

> library(pwr)
> pwr.t.test(d=4/sd(redchol), sig.level=0.05, power=0.9, type=✬one.sample✬)

One-sample t test power calculation

n = 74.31763
d = 0.3810377
sig.level = 0.05
power = 0.9
alternative = two.sided

Solution: We require a sample of size at least 75. 1 mark for t-test, 1 mark for some right
arguments, 1 mark for correct answer.

Downloaded by Zihan Yan ([email protected])