Seminar completed 23 October 2025
View slides
View recording

Presenter: Maeve McGillycuddy, Statistical Consultant, UNSW Stats Central 

Overview: Many studies collect multiple, often correlated outcome measures, also known as multivariate data, such as responses to items within a psychological scale, abundance of multiple species across sites, or concentrations of chemical elements. Researchers often struggle with deciding how to analyse such data appropriately. This seminar will explore analytical strategies for analysing multiple correlated outcomes in low-dimensional settings. Using examples, we will illustrate different approaches, and discuss the trade-offs between statistical power, interpretability and model complexity.

Seminar completed 25 September 2025
View slides
View recording

Presenter: Eve Slavich, Statistical Consultant, UNSW Stats Central

Overview: A pervasive misinterpretation in research occurs when investigators conclude that two treatments or conditions are "the same" based solely on obtaining a non-significant p-value. This common error stems from a fundamental misunderstanding: the absence of evidence for a difference is not evidence of equivalence. Traditional hypothesis testing is designed to detect differences, not to demonstrate similarity. This presentation is designed for researchers across disciplines who regularly encounter questions about whether treatments, methods, or conditions can be considered "equivalent".

Seminar completed 28 August 2025
View slides
View recording

Presenter: Tony Zhang, Statistical Consultant, UNSW Stats Central

Overview: Latent class analysis (LCA) and latent profile analysis (LPA) are powerful tools for uncovering unobserved subgroups within complex health data. In this talk, I will explore how these methods are applied in medical research, using examples from identifying lung function trajectories in the general population to uncovering comorbidity clusters in specific patient groups.

Seminar completed 24 July 2025
View slides
View recording

Presenter: Ben Maslen, Statistical Consultant, UNSW Stats Central

We will discuss methods to deal with non-linear data, with some examples in R using splines with generalised additive models.

Seminar completed 17 April 2025
View slides
View recording

Presenter: A/Professor Nancy Briggs, Statistical Consultant, UNSW Stats Central

Longitudinal studies often have missing data due to participants being lost to follow up. We often consider these missing data ignorable (Missing Completely At Random or Missing At Random) and analyse them using methods that produce unbiased estimates of the effect we are assessing. However, there are often good reasons why the missing data are not ignorable.

This seminar will demonstrate one approach to estimating a treatment effect in longitudinal data with data that are Missing Not At Random.

Seminar completed 27 March 2025
View slides
View recording

Presenter: Luz Palacios-Derflingher, Biostatistician, Statistical Consultant, UNSW Stats Central

Despite the American Statistical Association releasing a statement almost 10 years ago saying that "Scientific conclusions and business or policy decisions should not be based only on whether a p-value passes a specific threshold", in practice, still some conclusions and decisions are being based on a single value (usually a significance level of 0.05).The talk will cover the importance of confidence intervals in making conclusions, rather than relying only on p-values. It will include examples of confidence interval performance.

Seminar completed 27 February 2025
View slides
View recording
Presenter: Nickson Ning, Statistical Consultant, UNSW Stats Central

In experiments that aim to determine whether a treatment has an effect or not, randomised controlled trials (RCTs) are considered a gold standard. When RCTs are not possible due to practical or ethical constraints, observational data is often relied on instead. However, observational data is often wrought with problems such as confounding, which motivates the use of various causal inference methods. In this seminar, we begin by introducing and defining causal effects using Rubin’s potential outcomes framework. We then compare RCTs and observational data, and explore three causal inference methods: G-computation, Inverse Probability of Treatment Weighting, and Targeted Maximum Likelihood Estimation. We demonstrate how these methods address confounding, and also discuss their limitations.