Available in 2024
Course code

BIOS6940

Units

10 units

Level

6000 level

Course handbook

Description

This course presents methods for analyses of categorical data, with a key focus on generalised linear models (GLMs). Students will study contingency tables and gain experience with a range of generalised linear models appropriate to binary, count, nominal, ordinal and time-to-event response variables. There will be an emphasis on understanding model equations, fitting statistical models using statistical software, using fitted model equations to identify estimates of interest, checking model assumptions and accurately interpreting results for reporting or publication.


Availability2024 Course Timetables

Online

  • Semester 1 - 2024

Learning outcomes

On successful completion of the course students will be able to:

1. Construct and analyse contingency tables for a range of study designs, using both manual calculations and statistical software

2. Determine when contingency table analysis requires methods for stratified or matched data, and correctly apply and appraise the results of such methods

3. Identify and define the components of a generalised linear model (GLM)

4. For a given generalised linear model equation and sample from an exponential family distribution, estimate distributional parameters and interpret these in relation to a linear predictor

5. Propose and fit appropriate GLMs for binary, ordinal and nominal data, and interpret results in relation to the original research question

6. Construct GLMs for count data using Poisson or negative binomial regression and defend the choice of model

7. Recognise time-to-event response variables and correctly format and analyse these using Kaplan-Meier estimators and Cox regression

8. Present results from GLMs in publication-ready tables and interpret results for a non-statistical audience


Content

Topics covered include contingency tables, the exponential family and generalised linear models. Students will learn about estimation and modelling using logistic regression for binary, ordinal and multinomial responses, Poisson and negative binomial regression for counts, and Cox regression for time to event responses.  There will be an emphasis on fitting appropriate models using Stata or SAS statistical software, checking model fit and interpreting results.


Assumed knowledge

Students are assumed to have basic algebraic skills, including familiarity with mathematical functions and equations and the ability to rearrange equations to solve for terms of interest. It is also assumed that students understand the relationship between logarithmic and exponential functions and can convert from one form to the other. Students are also assumed to have a basic understanding of probability, or at a minimum, not be daunted by exposure to probabilistic concepts. The course refers to various probability distributions (e.g., normal, binomial, Poisson distributions), to teach students how to identify the distribution(s) appropriate for a given response variable. Mathematical notation is used to describe distributions and represent the parameter(s) being estimated by a regression model, such as the mean of a normal distribution, or the probability of a binomial event. Use of such terminology/notation is largely descriptive and complex calculations are not required. Prior completion of a course in probability theory (e.g., BIOS6170) is not necessary to succeed in this course but will enhance students’ theoretical understanding of the models. Students are also assumed to have basic, prior experience with simple and multiple regression, e.g., linear or logistic regression. This may be gained by prior completion of an introductory regression course, e.g., BIOS6920 Biostatistics B or BIOS6070 Linear Regression, or similar practical experience. BIOS6940 is an intermediate regression course focused on providing a unified approach to diverse regression methods. Students lacking a basic understanding of regression methods may find this course challenging.


Assessment items

Written Assignment: Assignment 1

Written Assignment: Assignment 2

Written Assignment: Assignment 3

Written Assignment: Assignment 4


Contact hours

Semester 1 - 2024 - Online

Online Activity-1
  • Online 10 hour(s) per week(s) for 13 week(s) starting in week 1
  • Contact Hours are an Indication Only

Course outline