SSA ACT branch invites everyone to attend our 2024 Dennis Trewin prize, featuring three outstanding early career researchers from diverse backgrounds who will wowing us with their recent research projects in statistics!
The Dennis Trewin Prize, named after the former Australian Statistician Dennis Trewin AO, is awarded annually by the Canberra Branch for outstanding early career research in an area related to statistics and/or data science conducted within the ACT or regional NSW outside Newcastle-Sydney-Wollongong.
This year we have a short-list of three candidates. They will be presenting their work at our late October meeting and a selection panel will determine the winner. Each candidate will have 20-25 minutes for their presentation and 5 minutes for audience questions. In partnership with the Australian Bureau of Statistics, monetary prizes totalling $1,000 or more will be on offer and distributed at the discretion of the selection panel.
Date: Tuesday 29 October 2024
Time: Starts 5:15pm and finishes by 7:15pm
Venue: Room 4.05 in the Marie Reay Teaching Centre, the Australian National University, or via Zoom.
Zoom link: https://anu.zoom.us/j/82666579719?pwd=WuFm8X0JaA64dar36NdYVpd5x3YlMH.1
The zoom link will be open at 5.00pm or soon afterwards.
Speakers (not necessarily in this order; abstracts provided below)
1. Owen Forbes: clusterBMA: Bayesian model averaging for clustering
2. Houren Hong: Extrinsic Single-index model for Spherical Data and Its M-estimation
3. Elle Saber: Resemblance between microbiomes: the behaviour of distances between seemingly compositional data inspired by a Salmon Breeding Program.
Dinner:
After the talks we will be holding a dinner at Badger & Co in the adjoining building at ANU (Badger & Co – Uni Pub – ANU – Canberra (badgerandco.com.au)) at 7.30pm. If you are interested in attending the dinner, please let me know by 5pm Monday 28 October by entering your details at SSA Canberra Branch dinner attendance sheet, or contacting me (warren.muller@csiro.au; 0407 916 868). Please regard this as a firm commitment, not just an intention. For withdrawals after the deadline, please remove your name from the sheet and phone or text me (0407 916 868).
NOTE: We are offering discounts to SSA early career and student members who attend dinner! For this meeting, dinners will be a fixed charge of $5 for student members and $10 for early career members. As the venue is card payment only, subsidised participants should pay Warren Muller, who will pay for their meal by card. Other participants should purchase their own meals.
————
Owen Forbes : clusterBMA: Bayesian model averaging for clustering
Abstract: clusterBMA is a novel framework that can probabilistically combine clustering solutions from multiple algorithms, addressing uncertainty in unsupervised clustering through model averaging. This method generates a weighted average across different clustering algorithms by using cluster internal validation indices to approximate the posterior model probability, followed by symmetric simplex matrix factorisation to compute combined probabilistic cluster allocations. In simulation studies we demonstrate clusterBMA's superior performance compared to other ensemble clustering methods, particularly for high-dimensional data with low cluster separation. An applied case study using electroencephalography (EEG) data showcases the method's practical utility in identifying probabilistic clusters of individuals, highlighting its potential for clinical applications and statistical communication in health data analysis.
Houren Hong : Extrinsic Single-index model for Spherical Data and Its M-estimation
Abstract: While many intrinsic approaches have developed for the regression analysis of spherical data, there is a notable lack of studies on extrinsic regression models. We propose an Extrinsic Single-Index Model (ESIM), which, to our knowledge, is the first semiparametric model within the extrinsic framework. A key focus of this study is the M-estimation of ESIM, ensuring robust parameter estimation in the presence of outliers or model misspecification. We establish the asymptotic distributions of both the parametric and nonparametric components, and importantly, we formulate the influence function and standardized influence function to access the robustness of our approach.
Elle Saber : Resemblance between microbiomes: the behaviour of distances between seemingly compositional data inspired by a Salmon Breeding Program.
Abstract: This research addresses the question of how to estimate heritability of the microbiome as a phenotype. The premise of this work is ‘if a trait is heritable, then two individuals who are closely related should be more similar than two unrelated individuals’ - hence a first step to quantifying heritability is to measure similarity between individuals’ microbiomes. I examine the properties of popular distance measures on seemingly-compositional data and how this choice impacts the biological interpretation of the problem and find that the widely advocated Aitchison Distance behaves in a manner contradictory to most sensible biological interpretation, while other methods not deemed “compositionally valid” seem to find a better balance between statistical rigour with biological relevance.
@StatSocAus
