Project 2 2021

This document provides instructions for a data analysis project on sex differences in the condition of Tasmanian devils infected with a transmissible cancer. Students are asked to analyze a dataset containing information on 111 infected devils, including tumor volume, body condition, sex, and location. The analysis involves exploring relationships between predictors and body condition, performing backwards stepwise regression to identify significant predictors, predicting body conditions, and evaluating how well the model fits the data.

Uploaded by

Jason Zeng

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

24 views

Project 2 2021

Uploaded by

Jason Zeng

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 2

U NIVERSITY OF TASMANIA , S CHOOL OF N ATURAL S CIENCES (M ATHEMATICS )

KMA353 Data Handling and Statistics 3

Project 2
Due: 10am Monday March 29, 2021

Dr Shane A. Richards

S EX DIFFERENCES IN TASMANIAN D EVIL CONDITION WHEN

CANCEROUS ?

This project involves you analysing data collected by Ruiz-Aravena et al. (2018). Note that
much of the text describing the study is taken from this paper.
Tasmanian devil (Sarcophilus harrisii) populations have declined precipitously owing to
infection by a transmissible cancer (devil facial tumour disease, DFTD) that emerged in the
late 1990s. Since the emergence of DFTD, and as the disease spreads across Tasmania, the
number of devils has dropped up to 90% across 80% of the species’s distributional range. As
a result, the disease is expected to act as a strong selective force on hosts to develop mecha-
nisms of tolerance and/or resistance to the infection. The objective of the study is to look for
evidence for the ability of infected devils to cope with infection, which translates into host tol-
erance to the cancer. Specifically, the study aims to look for differences in tolerance between
males and females.
Devil populations were sampled 10 times at two sites in northern Tasmania (here labelled
as site A and site B) between February 2015 and August 2016. The total volume of tumour
tissue was estimated for individuals showing clinical signs of DFTD. The body condition of
each animal was also estimated and it is based on the relationship between body weight and
body size. In brief, body condition is the change in the weight of an animal relative to its
expected weight if it were a healthy individual. A value of zero indicates the weight of the
animal is comparable to a healthy animal having the same head width. A negative value
means the animal weighs less than expected, and vice-versa. More details on how the data
were collected can be found in Ruiz-Aravena et al. (2018).
The file DevilCondition.csv provides all the data needed to complete this assignment.
Each row corresponds to an infected animal. There are data for 111 animals. The data anal-
ysed by Ruiz-Aravena et al. (2018) included multiple observations of the same animal. In
order to simplify the analysis I have only included data taken from the first time an animal
was captured. The date of animal capture has also been removed to keep the analysis simple.
There are five named columns in this .csv file:

ID A unique identifier for each animal.

Sex Animal sex: M = male, F = female.

Site Site where the animal was captured: A, B.

1
LTV Log (base 10) of the volume of the tumour. For example, a spherical tumour with radius
10 mm would have a volume of 4188.8 mm3 , which is 3.62 in log base 10.

BC Body condition of the animal, defined as the weight of the animal relative to the mean
weight of a healthy animal of the same size. For example, a value of -0.25 implies the
animal weights 25% less than a healthy animal of comparable size.

Perform the following tasks using the Rmarkdown file Project_2.Rmd that is provided on
MyLO. Place your R code and text in the appropriate sections indicated below.

PART 1. D ATA EXPLORATION

1. Read the csv file into a data frame (or tibble) and make sure that the Sex and Site
columns are factors.

2. Use ggplot to produce a figure that indicates whether or not animal body condition is
related to the volume of the tumour (LTV), the sex of the animal, and the site where the
animal was captured.

3. Under the figure provide some clear and concise text that would be appropriate for a
figure legend.

4. By referring to the figure write some brief text discussing which of the three predic-
tor variables: LTV, Sex, or Site, you might expect to be statistically significant when
predicting BC. Include in your discussion whether or not you expect there to be a sta-
tistically significant interaction between LTV and Sex.

PART 2. B ACKWARDS S TEPWISE R EGRESSION

In this part of the assignment you will perform backwards step-wise regression (BSR) to de-
termine which of the three predictor variables (Sex, Site, TLV) are statistically significant
when predicting animal body condition.

1. Starting with the most complex model that includes all pairwise interactions and the
three-way interaction, perform a BSR analysis. Show each step of the analysis and
briefly explain which model term was removed during each step.

2. Show the estimated model parameters for the final selected model and write some brief
text explaining what effect each of the predictors has on body condition.

3. Use the final selected model to predict the expected body condition for each animal.

4. Plot the observed and predicted body conditions.

5. Using the plot above and possibly other informative plots briefly explain if you think
the final model adequately describes the observed data. Explain your reasoning.

Make sure you attempt to submit a single PDF document to MyLO by the due date and time. For
this project I am not expecting a formal report but you must perform all tasks set and answer
all questions clearly and concisely.

R EFERENCE
Ruiz-Aravena M, Jones ME, Carver S, Estay S, Espejo C, Storfer A, Hamede RK. 2018 Sex bias
in ability to cope with cancer: Tasmanian devils and facial tumour disease. Proc. R. Soc. B
285: 20182239. http://dx.doi.org/10.1098/rspb.2018.2239