Statistics Across Campuses

  Friday 2 June, 4 PM Social to follow, 5 - 6 PM Scott Hottovy, Associate Professor of Mathematics, United States Naval Academy Hybrid: Red Centre 4082 https://unsw.zoom.us/j/88495626621 “Convergence of Rain Process Models” Abstract:  Representing rain in global climate models continues to be a challenge. Currently, models generally rain too often and too little. Additionally, the models have trouble capturing variability in rain data. One possible solution to increasing variability in a model is to use a stochastic process. A variety of stochastic models have been used to describe time series of precipitation or rainfall. Since many of these stochastic models are simplistic, it is desirable to develop connections between the stochastic models and the underlying physics of rain. In this talk, I will describe simple models of rain in a single column model as a stochastic differential equation (SDE) with a switch. The inclusion of this switch leads to a model with hysteres...

  Stochastic flows in infinite dimensions Date: 25 May 2023, Thursday Time: 5pm AEST Statistics and Stochastic colloquium (part of the Colloquium Series of the Department of Mathematics and Statistics) at La Trobe University jointly organized with the Probability Victoria Seminar. Contact the organizers: Andriy Olenko a.olenko@latrobe.edu.au, Kostya Borovkov kostya.borovkov@gmail.com Speaker: Prof Ben Goldys (The University of Sydney, the Commonwealth of Australia) Abstract: Stochastic flows associated with finite dimensional SDEs and random dynamical systems are an important tool for the study of large time behaviour of solutions and for their pathwise analysis. The existence of (smooth) stochastic flows in finite dimensions is well understood, see for example books by Kunita. The famous Skorokhod example shows that the situation is completely different in infinite dimensions (hence for stochastic PDEs). The existence of the flow for infinite dimensional stochastic syst...

Bespoke Realized Volatility: Tailored Measures of Risk for Volatility Prediction Speaker:  Prof. Andrew Patton   (Duke University) Date/Time:   Wednesday 10 May, 2 pm -3 pm Location:   110 Finance Decision Lab, 4 Eastern Rd / Online  via Zoom Abstract:  Standard realized volatility (RV) measures estimate the latent volatility of an asset price using high frequency  data  with  no  reference  to  how  or  where  the  estimate  will  subsequently be used. This paper presents methods for “tailoring” the estimate of volatilityto the application in which it will be used. For example,  if the volatility measure willbe  used  in  a  specific  parametric  forecasting  model,  it may  be possible  to  exploit  that information and construct a better measure of volatility.  We use methods from machine learning to estimate optimal “bespoke”...

  Geodesic random walks, diffusion processes, and Brownian motion on Finsler manifolds Date: 11 May 2023, Thursday Time: 5pm AEST Statistics and Stochastic colloquium (part of the Colloquium Series of the Department of Mathematics and Statistics) at La Trobe University jointly organized with the Probability Victoria Seminar. Contact the organizers: Andriy Olenko a.olenko@latrobe.edu.au, Kostya Borovkov kostya.borovkov@gmail.com Speaker: Prof Vladimir S. Matveev (Die Friedrich-Schiller-Universität Jena, the Federal Republic of Germany // La Trobe University, the Commonwealth of Australia) Abstract: We show that geodesic random walks on a complete Finsler manifold of bounded geometry converge to a diffusion process which is, up to a drift, the Brownian motion corresponding to a Riemannian metric. In particular, the generator of the limit process is a non-degenerate elliptic second-order partial differential operator for which we give a precise integral formula. If the geodes...

  Friday 5 May, 4 PM AET Doan Khue Dung (KD) Dang, Lecturer, School of Mathematics and Statistics, University of Melbourne Virtual Zoom link: https://unsw.zoom.us/j/88495626621 “Bayesian structural equation modelling: methods and applications.” Abstract: Structural equation models (SEMs) are commonly used in social and behavioural sciences to study the structural relationships between observed variables and latent constructs or unobserved variables. Recently, Bayesian fitting procedures for SEMs have received more attention, as they overcome the issues of frequentist approaches when the number of observations is small and facilitate the adoption of more flexible model structures. In this talk, I will present an application of Bayesian SEMs in analysing the nature of the dose-response relation between prenatal alcohol exposure and child cognition, based on data from multiple cohorts. I will then introduce a framework for fitting Bayesian SEMs via variational approximations, th...