Written Assignment Part B
For this assignment, you will look at an adaptation of a
east cancer dataset
(AssignmentB_data.txt). There are 686 patients in this dataset, with the variables
available (two of which are continuous, two binary and one factor) described below,
including overall survival and overall survival status.
The variables are:
Patient Patient ID PT1 – 686
Age Age at Diagnosis Years
MenopauseStatus Menopausal Status 1 = No, 2 = Yes
HormoneTherapy Hormone Therapy 1 = No, 2 = Yes
TumourSize Tumor Size mm
TumourGrade Tumor Grade 1 – 3
SurvTime Time to Death Days
SurvStat Death Censoring 0 = Censored, 1 = Death
Objective: To analyse (in R) the variables affecting overall survival in this group of
patients using the Cox Proportional Hazards model. This will include the following
tasks:
Task 1: Summarization of data [6 marks]
Task 2: Checking of Proportional Hazards and linearity (where appropriate)
assumptions [6 marks]
Task 3: Univariate analysis [6 marks]
Task 4: Multivariate analysis [6 marks]
Task 5: Variable selection [3 marks]
Task 6: Final multivariate model [4 marks]
Task 7: Interpretation [6 marks]
Task 8: R code [6 marks]
Task 9: Presentation [7 marks]
Report: You should provide a report detailing the data composition, methods used,
your interpretation of the results. The report should include tables and plots (including
table and figure titles) where appropriate. There is no maximum nor minimum word
limits for this assignment, though a report of around two sides of A4 (excluding tables,
figures) would be sufficient. The final report must be in English and in .docx (or Open
Source equivalent) or in .PDF format (preferably). The R code that you use MUST
also be included in a separate R script file (.R) to enable the marker to test run your
code. Please pack your assignment work (the .docx or .pdf report and the .R script)
into a single zip file, and submit it to the submission dropbox provided for this
assignment.
(Use examples given in Week 3 and in particular Week 4 for presenting and laying-
out of your work)
NOTE: you can assume a variable has met the PH assumption if p > 0.05 in the
appropriate test. You do not need to test for interactions or associations between
variables in this assignment. You are not expected to provide a clinical
interpretation of the results.
Source: Description of the variables in the German Breast Cancer Study (GBCS) dataset shown in
Table 1.4 of Hosmer, D.W. and Lemeshow, S. and May, S XXXXXXXXXXApplied Survival Analysis:
Regression Modeling of Time to Event Data: Second Edition, John Wiley and Sons Inc., New York, NY.
Obtained from: ftp:
ftp.wiley.com/public/sci_tech_med/survival/. Used with adaptations.