The Statistical Consulting Centre presents a free monthly seminar series on topics of interest to analysts and researchers including graduate students.
The seminars will cover a wide range of topics, with a practical and applied focus. Each presentation will run for around 30 minutes, allowing plenty of time for questions. The seminars are run online on the second Friday of each month via Zoom, starting at 12:30pm.
Get updates about the SCC seminar series
Meta-analysis: some surprising statistical challenges
12:30 pm Friday 11 September 2026
Meta-analysis is a powerful statistical technique for combining results across studies identified in a systematic review. There are useful frameworks and guidelines for carrying out and reporting meta-analyses, such as those provided by the Cochrane Collaboration.
This introductory talk will discuss some of the statistical challenges that arise once the studies for meta-analysis have been identified and the appropriate data need to be extracted. This stage in the meta-analysis is not always as straightforward as it first appears. This seminar will provide examples of common problems in deriving the data needed for the analysis and illustrate some solutions.
Sue Finch is Deputy Director of the Statistical Consulting Centre. Despite being a statistician, she’s lost count of the number of meta-analyses she has carried out over the past 25 years in supporting PhD students, academics and clients making submissions to the Pharmaceutical Benefits Advisory Board.
Branch out! What tree-based methods can do for you
12:30 pm Friday 9 October 2026
Dr Jeremy Silver
Data science methods based on decision trees can be a useful in many contexts, such as classification, regression and prediction tasks. Decision trees can be used for presentation or communication of how predictors flow through to inference about an outcome. Ensembles of decision trees can be used in a range of machine-learning tasks; such methods can capture interactions between predictors and non-linear predictor-response relationships, are often very accurate, are compatible with large datasets, and have inherent variable-selection capabilities. In this talk, I will give an introduction and an overview of these methods, including some examples.
Stayin’ Alive – Survival analysis: the basics and some interesting wrinkles
12:30 pm Friday 13 November 2026
Professor Ian Gordon
Survival analysis deals with an outcome that is the time until a key event, from a suitable starting point. The archetypal context is human life, where death is the event. We are interested in studying the pattern of mortality, and what may influence the death rate. As in any interesting statistical context, there are research questions about the influence of explanatory variables.
The event could be of a different nature, and could even be something desirable, such as “resolution of symptoms”.
This talk will discuss the standard approaches used to analyse such data, which usually have to deal with “censoring” of some sort. This is what makes survival analysis special.
I will also outline a number of quirks and traps for the unwary: aspects of survival data and analysis that need attention. These include estimation of the mean survival, definitions of the time origin, defining groups on the basis of outcomes, composite outcomes, the “immortality bias”, interval censoring, and more.
Upcoming seminars
- More to come in 2027!
Missed a seminar? We can present to your group
If you would like us to present one of the seminars listed below to your lab or research group, contact Professor Ian Gordon.
-
Linear models are the foundation for most classical methods in statistics; they are simple, powerful and readily interpretable. However, the linearity assumption does not always hold, and it can be useful to know how to relax this assumption. You can come to the wrong conclusion if you have "missed a squiggle" without knowing it! We will explore generalised additive models as a means of testing the linearity assumption and adapting the model using patterns in the data themselves. We will discuss how you can get the most out of such models, and the trade-offs involved in moving away from the well-worn path of linear models.
-
Data with a hierarchical structure is very common in many fields. Terms used to describe this kind of data include multi-level data, longitudinal data, split-plot designs or repeated measures designs. These terms arose from different contexts, but they share a common feature. Observations are not independent, and there is structure to the dependence which can be incorporated into a statistical model.
Linear mixed-effects models ("mixed models") are a commonly used statistical method for multi-level or longitudinal data. Mixed models are versatile, capable of accounting for complex hierarchical structures, missing outcome data, and outcome variables with different distributions.
-
It is generally true that the more hypothesis tests you perform, the greater the probability that one of those tests is statistically significant by chance, the so-called multiple testing problem. At the Statistical Consulting Centre, we regularly receive enquiries regarding this issue. Reviewers often ask for adjustment for multiple comparisons. But how important is it? What can you do about it?
This seminar will outline some principles to consider, along with practical advice.
-
Researchers have access to a wide range of statistical software, but it can sometimes be challenging to decide which is the best choice to suit particular needs. This seminar will showcase some popular statistical software - R, SPSS and Minitab. We will discuss the considerations that can guide the choice of software based on individual preferences, needs and longer term goals.
-
“You can’t fix by analysis what you bungled by design.” (Light, Singer and Willett, 1990)
Good study design in one of the foundations of meaningful empirical research. In this seminar, we discuss the principles of good experimental design. These principles are important for anyone contemplating running an experiment.
-
You’re a researcher at the University of Melbourne. Or a member of an analytics team in the finance industry. Or a market researcher.
Do you need help from statistics, or data science? Or something else again?
In this seminar we discuss these terms, and how you can navigate your way through the sometimes bewildering array of concepts and activities that aim to support or collaborate with anyone pursuing quantitative inquiry.