BEGIN:VCALENDAR
VERSION:2.0
PRODID:researchseminars.org
CALSCALE:GREGORIAN
X-WR-CALNAME:researchseminars.org
BEGIN:VEVENT
SUMMARY:Prof Robert Griffiths (Monash University)
DTSTART;VALUE=DATE-TIME:20200430T010000Z
DTEND;VALUE=DATE-TIME:20200430T020000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/1
DESCRIPTION:Title: Lambda coalescent trees and graphs\nby Prof Robert Gri
ffiths (Monash University) as part of La Trobe University Statistics and S
tochastic zoom seminar\n\n\nAbstract\nThe Lambda coalescent introduced by
Pitman (1999) and Sagitov (1999) is a random tree which has multiple merg
ers. It is a dual to a Lambda-Fleming-Viot process which describes a popu
lation of individuals with births and deaths\, where a single individual's
children can contribute a large proportion of the population. The popula
tion process has jumps at times where individuals give birth. The Wright-
Fisher diffusion in contrast\, being a diffusion\, is continuous over tim
e. The Kingman coalescent\, a random binary tree\, is dual to the Wright
-Fisher diffusion.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/1/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Dr Jessica Kasza (Monash University)
DTSTART;VALUE=DATE-TIME:20200507T020000Z
DTEND;VALUE=DATE-TIME:20200507T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/2
DESCRIPTION:Title: Cross-overs\, stepped wedges and staircases: some recent w
ork in longitudinal cluster randomised trials\nby Dr Jessica Kasza (Mo
nash University) as part of La Trobe University Statistics and Stochastic
zoom seminar\n\n\nAbstract\nAlthough individually randomised trials are th
e gold standard for assessing the impact of new treatments on patient out
comes\, cluster randomised trials are necessary when testing the effect o
f healthcare provider-level changes on patient outcomes\, e.g. the effect
of a hospital-wide handwashing program on the number of patients who acq
uire infections in hospital. Cluster randomised trials often require larg
e numbers of clusters and thus can be infeasible\, but longitudinal cluste
r randomised trials\, where clusters may switch between intervention and c
ontrol\, require smaller sample sizes. Cross-overs\, stepped wedges and s
taircases are all particular variants of longitudinal cluster randomised
trials that are being conducted with increasing frequency. However\, many
of the underlying statistical aspects of these designs remain under-expl
ored.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/2/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Dr David Frazier (Monash University)
DTSTART;VALUE=DATE-TIME:20200611T020000Z
DTEND;VALUE=DATE-TIME:20200611T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/3
DESCRIPTION:Title: Robust and Efficient Approximate Bayesian Computation: A M
inimum Distance Approach\nby Dr David Frazier (Monash University) as p
art of La Trobe University Statistics and Stochastic zoom seminar\n\n\nAbs
tract\nIn many instances\, the application of approximate Bayesian methods
is hampered by two practical features: 1) the requirement to project the
data down to low-dimensional summary\, including the choice of this projec
tion\, and which ultimately yields inefficient inference\; 2) a possible l
ack of robustness of these methods to deviations from the underlying model
structure. Motivated by these efficiency and robustness concerns\, we con
struct a new Bayesian method that can deliver efficient estimators when th
e underlying model is well-specified\, and which is simultaneously robust
to certain forms of model misspecification. This new approach bypasses the
calculation of summaries by considering a norm between empirical and simu
lated probability measures. For specific choices of the norm\, we demonstr
ate that this approach can be as efficient as exact Bayesian inference\, a
nd is robust to deviations from the underlying model assumptions. We illus
trate this approach using several examples that have featured in the liter
ature.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/3/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Dr Damjan Vukcevic (University of Melbourne)
DTSTART;VALUE=DATE-TIME:20200618T020000Z
DTEND;VALUE=DATE-TIME:20200618T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/4
DESCRIPTION:Title: Analysis of repeated categorical ratings: going beyond int
er-rater agreement.\nby Dr Damjan Vukcevic (University of Melbourne) a
s part of La Trobe University Statistics and Stochastic zoom seminar\n\n\n
Abstract\nA common task in health and medicine is the classification of pa
tient information into one of several categories by a trained expert. This
could include assessing the presence and type of a tumour from a medical
image or providing a disease diagnosis from a series of medical tests. Oft
en such judgements are hard to make and error prone: two experts may rate
the same scenario differently or the same expert may provide alternative r
atings of the same scenario when rating it multiple times on different occ
asions.\n\nAnalysing the performance of such expert ‘raters’\, and the
accuracy of their ‘ratings’ across a series of ‘items’\, is a com
mon theme in much of the health and medical literature\, especially in the
setting where the true underlying category is unknown. Existing approache
s\, such as Cohen’s kappa\, focus only on assessing inter-agreement\, an
d have known problems stemming from the lack of any notion of underlying t
ruth and the difficulty of coping with repeated ratings by the same rater.
\n\nHere we present and implement methods that explicitly model an underly
ing true category for each item and can cope naturally with any number of
ratings for each item\, including repeated ratings by the same rater. We i
mplement Bayesian versions of these models using the probabilistic program
ming language Stan\, and create an R package to fit and interrogate the ou
tput of these models.\n\nUsing real and simulated datasets\, which are des
igned to mimic a wide range of medical scenarios\, we test the performance
of these models in estimating the true class of each item. We also explor
e situations such as having raters with much poorer accuracy\, and compari
sons with other (non-model-based) approaches.\n\nA common task in health a
nd medicine is the classification of patient information into one of sever
al categories by a trained expert. This could include assessing the presen
ce and type of a tumour from a medical image or providing a disease diagno
sis from a series of medical tests. Often such judgements are hard to make
and error prone: two experts may rate the same scenario differently or th
e same expert may provide alternative ratings of the same scenario when ra
ting it multiple times on different occasions.\n\nAnalysing the performanc
e of such expert ‘raters’\, and the accuracy of their ‘ratings’ ac
ross a series of ‘items’\, is a common theme in much of the health and
medical literature\, especially in the setting where the true underlying
category is unknown. Existing approaches\, such as Cohen’s kappa\, focus
only on assessing inter-agreement\, and have known problems stemming from
the lack of any notion of underlying truth and the difficulty of coping w
ith repeated ratings by the same rater.\n\nHere we present and implement m
ethods that explicitly model an underlying true category for each item and
can cope naturally with any number of ratings for each item\, including r
epeated ratings by the same rater. We implement Bayesian versions of these
models using the probabilistic programming language Stan\, and create an
R package to fit and interrogate the output of these models.\n\nUsing real
and simulated datasets\, which are designed to mimic a wide range of medi
cal scenarios\, we test the performance of these models in estimating the
true class of each item. We also explore situations such as having raters
with much poorer accuracy\, and comparisons with other (non-model-based) a
pproaches.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/4/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Dr Mumtaz Hussain (La Trobe University)
DTSTART;VALUE=DATE-TIME:20200625T020000Z
DTEND;VALUE=DATE-TIME:20200625T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/5
DESCRIPTION:Title: Metric number theory via geometry and dynamics: Mahler to
Margulis\nby Dr Mumtaz Hussain (La Trobe University) as part of La Tro
be University Statistics and Stochastic zoom seminar\n\n\nAbstract\nThere
are two well-known approaches in solving the measure theoretic problems in
Diophantine approximation. The metrical approach arise from the geometry
of numbers and the ergodic theoretic approach arise from the dynamics on
the space of lattices. One of the main ingredients in the geometry of numb
ers is the usage of Borel-Cantelli lemmas from probability theory. Dynamic
s on the space of lattices rely on the Dani correspondence principle (1985
) which was extensively developed further by Margulis and Kleinbock. I w
ill discuss both of these approaches and along the way discuss some well-k
nown results such as the resolutions of Oppenheim (1929)\, Mahler (1932) a
nd Sprindzuk (1965) conjectures which influenced my research in the last
few years.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/5/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Marina Masioti (La Trobe University)
DTSTART;VALUE=DATE-TIME:20200603T010000Z
DTEND;VALUE=DATE-TIME:20200603T020000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/6
DESCRIPTION:Title: Optimal transformations for dimension reduction and the pr
oblem of eigenvalue switching\nby Marina Masioti (La Trobe University)
as part of La Trobe University Statistics and Stochastic zoom seminar\n\n
Abstract: TBA\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/6/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Jason Gavin Grealey (La Trobe University)
DTSTART;VALUE=DATE-TIME:20200319T040000Z
DTEND;VALUE=DATE-TIME:20200319T050000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/7
DESCRIPTION:Title: Investigating the utility of neural networks in genomic pr
ediction\nby Jason Gavin Grealey (La Trobe University) as part of La T
robe University Statistics and Stochastic zoom seminar\n\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/7/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Illia Donhauzer (La Trobe University)
DTSTART;VALUE=DATE-TIME:20200904T020000Z
DTEND;VALUE=DATE-TIME:20200904T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/8
DESCRIPTION:Title: Asymptotic behaviour of functionals of random fields\n
by Illia Donhauzer (La Trobe University) as part of La Trobe University St
atistics and Stochastic zoom seminar\n\n\nAbstract\nThe talk is about the
asymptotic behaviour of functionals of long-range dependent random fields.
The Strong Law of Large Numbers (SLLN) and new properties of the limit pr
ocesses in the Non-central Limit Theorem (NLT) will be discussed.\n\nThe S
LLN for integral functionals of random fields with unboundedly increasing
covariances will be presented. The SLLN is derived for the case of increas
ing domains. Conditions on covariance functions such that the SLLN holds w
ill be provided. The considered scenarios include non-stationary random fi
elds. The discussion about applications to weak and long-range dependent r
andom fields and numerical examples will be shown.\n\nNew properties of ge
neralized Hermite-type processes that arise in NLT for integral functional
s of long-range dependent random fields will be demonstrated. Contrary to
the classical one-dimensional case\, it will be shown that for any choice
of a multidimensional observation window the generalized Hermite-type proc
ess has non-stationary increments.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/8/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Dr Jie Yen Fan (Monash University)
DTSTART;VALUE=DATE-TIME:20201001T020000Z
DTEND;VALUE=DATE-TIME:20201001T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/9
DESCRIPTION:Title: Multi-type age-structured population model\nby Dr Jie
Yen Fan (Monash University) as part of La Trobe University Statistics and
Stochastic zoom seminar\n\n\nAbstract\nPopulation process in general setti
ng\, where each individual reproduces and dies depending on the state (suc
h as age and type) of the individual as well as the entire population\, of
fers a more realistic framework to population modelling. Formulating the p
opulation dynamics as a measure-valued stochastic process allows us to inc
orporate such dependence. We describe the dynamics of a multi-type age-str
uctured population as a measure-valued process\, and obtain its asymptotic
s\, in particular\, the law of large numbers and the central limit theorem
.\n\nJoint work with Kais Hamza\, Peter Jagers and Fima Klebaner.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/9/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Mason Terrett (La Trobe University)
DTSTART;VALUE=DATE-TIME:20201001T030500Z
DTEND;VALUE=DATE-TIME:20201001T040000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/10
DESCRIPTION:Title: SARGDV: Efficient identification of groundwater-dependent
vegetation using synthetic aperture radar\nby Mason Terrett (La Trobe
University) as part of La Trobe University Statistics and Stochastic zoom
seminar\n\n\nAbstract\nGroundwater depletion impacts the sustainability o
f numerous groundwater-dependent vegetation (GDV) globally\, placing signi
ficant stress on their capacity to provide environmental and ecological su
pport for flora\, fauna\, and anthropic benefits. Cost effective methods o
f GDV identification will enable strategic protection of these critical ec
ological systems\, through improved and sustainable groundwater management
by communities and industry. Recent application of synthetic aperture rad
ar (SAR) earth observation data in Australia has demonstrated the utility
of radar for identifying terrestrial groundwater-dependent ecosystems at s
cale. Our research included the development of SARGDV\, a binary classific
ation model\, which uses the extreme gradient boosting (XGBoost) algorithm
in conjunction with three data cubes composed of Sentinel-1 SAR interfero
metric wide images. Our method may be used to support the protection of GD
V communities globally by providing a long term\, cost-effective solution
to identify GDVs over variable regions and climates\, via the use of freel
y available\, high-resolution\, globally available Sentinel-1 SAR data set
s. Our method offers global water management agencies a means toward more
sustainable management of regional groundwater resources by providing an e
fficient method to identify significant GDV occurrence within areas where
substantial groundwater extraction is ongoing.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/10/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Dr Tingjin Chu (University of Melbourne)
DTSTART;VALUE=DATE-TIME:20201015T010000Z
DTEND;VALUE=DATE-TIME:20201015T020000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/11
DESCRIPTION:Title: Large Spatial Data Modeling and Analysis: A Krylov Subspa
ce Approach\nby Dr Tingjin Chu (University of Melbourne) as part of La
Trobe University Statistics and Stochastic zoom seminar\n\n\nAbstract\nEs
timating the parameters of spatial models for large spatial datasets can b
e computationally challenging\, as it involves repeated evaluation of siza
ble spatial covariance matrices. In this paper\, we aim to develop Krylov
subspace based methods that are computationally efficient for large spatia
l data. Specifically\, we approximate the inverse and the log-determinant
of the spatial covariance matrix in the log-likelihood function via conjug
ate gradient and stochastic Lanczos on a Krylov subspace. These methods re
duce the computational complexity from $O(N^3)$ to $O(N^2)$ and $O(N\\log
N)$ for dense and sparse matrices\, respectively. Moreover\, we quantify t
he difference between the approximated log-likelihood function and the ori
ginal log-likelihood function and establish the consistency of parameter e
stimates. Simulation studies are conducted to examine the computational e
fficiency as well as the finite-sample properties. For illustration\, our
methodology is applied to analyze a large LiDAR dataset.\n\nThis is joint
work with Jialuo Liu\, Jun Zhu and Haonan Wang.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/11/
END:VEVENT
BEGIN:VEVENT
SUMMARY:4 presenters (La Trobe University)
DTSTART;VALUE=DATE-TIME:20201105T230000Z
DTEND;VALUE=DATE-TIME:20201106T010000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/12
DESCRIPTION:Title: Theses talks by 4 students\nby 4 presenters (La Trobe
University) as part of La Trobe University Statistics and Stochastic zoom
seminar\n\n\nAbstract\nPresented by:\n\n10.00am Vibhooti Bhatnagar. A com
parison of AIC and nested t-tests for nested model selection.\n\n10.25am N
avdeep Kaur. Is corrected AIC really better than AIC?\n\n10.50am Satbir Ka
ur Bansal. Visualization of Variability of AIC.\n\n11.15 Ravindra Nath Dah
al. A review of Prediction Intervals obtained from model free machine lear
ning algorithms for point prediction\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/12/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Prof Gael M. Martin (Monash University)
DTSTART;VALUE=DATE-TIME:20201118T233000Z
DTEND;VALUE=DATE-TIME:20201119T003000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/13
DESCRIPTION:Title: Computing Bayes: Bayesian Computation from 1763 to the 21
st Century\nby Prof Gael M. Martin (Monash University) as part of La T
robe University Statistics and Stochastic zoom seminar\n\n\nAbstract\nThe
Bayesian statistical paradigm uses the language of probability to express
uncertainty about the phenomena that generate observed data. Probability d
istributions thus characterize Bayesian inference\, with the rules of prob
ability used to transform prior probability distributions for all unknowns
- models\, parameters\, latent variables - into posterior distributions\,
subsequent to the observation of data. Conducting Bayesian inference requ
ires the evaluation of integrals in which these probability distributions
appear. Bayesian computation is all about evaluating such integrals in the
typical case where no analytical solution exists. This paper takes the re
ader on a chronological tour of Bayesian computation over the past two and
a half centuries. Beginning with the one-dimensional integral first confr
onted by Bayes in 1763\, through to recent problems in which the unknowns
number in the millions\, we place all computational problems into a common
framework\, and describe all computational methods using a common notatio
n. The aim is to help new researchers in particular - and more generally t
hose interested in adopting a Bayesian approach to empirical work - make s
ense of the plethora of computational techniques that are now on offer\; u
nderstand when and why different methods are useful\; and see the links th
at do exist\, between them all.\n\nJoint results with David T. Frazier (Mo
nash University) and Christian P. Robert (University of Dauphine\, Paris).
The paper appears as an arXiv pre-print. We are revising it at the moment
\, but it won't change in its essence: https://arxiv.org/pdf/2004.06425.pd
f\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/13/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Ravindi Nanayakkara (La Trobe University)
DTSTART;VALUE=DATE-TIME:20201119T030000Z
DTEND;VALUE=DATE-TIME:20201119T040000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/14
DESCRIPTION:Title: Stochastic Modelling and Statistical Analysis of Dependen
t Data\nby Ravindi Nanayakkara (La Trobe University) as part of La Tro
be University Statistics and Stochastic zoom seminar\n\n\nAbstract\nFirst\
, we discuss the obtained results about the analysis of spherical monofrac
tal and multifractal random fields with cosmological applications. The Ré
nyi function plays an important role in the analysis of multifractal rando
m fields. For random fields on the sphere\, there are three models in the
literature where the Rényi function is known explicitly [1]. The main sta
tistical model used to describe CMB data in the literature is isotropic Ga
ussian fields. We present some new theoretical models\, numerical multifra
ctality studies and methodology based on simulating random fields\, comput
ing the Rényi function and the multifractal spectrum for different scenar
ios and actual CMB data. The results suggest that there may exist a very m
inor multifractality of the CMB data [2].\n\nNext\, we discuss the obtaine
d results about the asymptotic normality of simultaneous estimators of cyc
lic long-memory processes. Spectral singularities at non-zero frequencies
play an important role in investigating cyclic or seasonal time series. Th
e publication [3] introduced the generalized filtered method-of-moments ap
proach to simultaneously estimate singularity location and long-memory par
ameters. This study [4] continues investigations of these simultaneous est
imators. The results about asymptotic normality of several statistics are
obtained. The methodology includes wavelet transformations as a particular
case. The theoretical findings are illustrated by numerical results inclu
ding Meyer\, Shannon father wavelets and Mexican hat wavelets.\n\nFinally\
, we discuss multifractionality of spherical random fields with cosmologic
al applications. The Hölder exponent is used to measure the roughness in
a rigorous mathematical way [5]. In this study\, one dimensional and two d
imensional pointwise Hölder exponent values are computed for the CMB data
using the HEALPix ring ordering and nested ordering visualisations. The r
esults suggest that there exist a considerable multifractionality in CMB d
ata.\n\nReferences:\n\n Leonenko\, N. & Shieh\, N.R. (2013). Rényi fun
ction for multifractal random fields. Fractals\, 21(2)\, 1350009.\n Leo
nenko\, N.\, Nanayakkara\, R.\, & Olenko\, A. (2020). Analysis of Spherica
l Monofractal and Multifractal Random Fields. Stochastic Environmental Res
earch and Risk Assessment Journal. https://doi.org/10.1007/s00477-020-0191
1-z\n Alomari\, H. M.\, Ayache\, A.\, Fradon\, M. & Olenko\, A. (2020).
Estimation of cyclic long-memory parameters. Scandinavian Journal of Stat
istics\, 47(1) 104-133.\n Ayache\, A.\, Fradon\, M.\, Nanayakkara\, R.\
, & Olenko\, A. (2020). Asymptotic normality of simultaneous estimators of
cyclic long-memory processes. Submitted.\n Ayache\, A.\, & Véhel\, J.
L. (2004). On the identification of the pointwise Hölder exponent of the
generalized multifractional Brownian motion. Stochastic Processes and the
ir Applications\, 111(1)\, 119–56.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/14/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Nishika Ranathunga (La Trobe University)
DTSTART;VALUE=DATE-TIME:20201209T010000Z
DTEND;VALUE=DATE-TIME:20201209T020000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/15
DESCRIPTION:Title: Confidence Intervals in General Regression Models that Ut
ilize Uncertain Prior Information\nby Nishika Ranathunga (La Trobe Uni
versity) as part of La Trobe University Statistics and Stochastic zoom sem
inar\n\n\nAbstract\nWe consider a general regression model\, without a sca
le parameter. We construct a confidence interval for a scalar parameter of
interest that utilizes the uncertain prior information that a distinct sc
alar parameter takes the specified value. This confidence interval has goo
d coverage properties. It also has scaled expected length\, where the scal
ing is with respect to the usual confidence interval\, that is (a) substan
tially less than 1 when the prior information is correct\, (b) has a maxim
um value that is not too large and (c) is close to 1 when the data and pri
or information are highly discordant.\n\nFurthermore\, in Kabaila and Rana
thunga (2020)\, we solve the problem of numerically evaluating the expecte
d value of a smooth bounded function of a chi-distributed random variable\
, divided by the square root of the number of degrees of freedom\, using M
ori's transformation followed by the trapezoidal rule\, which is exponenti
ally convergent for suitable integrands. This problem arises in simultaneo
us inference\, selection and ranking of populations\, the evaluation of mu
ltivariate t probabilities and the assessment of coverage and expected vol
ume properties of non-standard confidence regions.\n\nWe apply this soluti
on in the R package ciuupi2 that computes the Kabaila and Giri (2009) conf
idence interval\, which utilizes the uncertain prior information in a line
ar regression model with unknown error variance. Previous computations of
this interval used MATLAB programs that were time-consuming to run. By wri
ting these programs in R\, the computation time is greatly reduced and the
y become freely available. We also assess a new definition of scaled expec
ted length.\n\nFinally\, we compare the computations of the log-likelihood
function for generalized linear mixed models using (a) adaptive Gauss-Her
mite quadrature and (b) importance sampling\, where both methods share the
same initial step (Kabaila and Ranathunga\, 2019).\n\nReferences:\n\n
Kabaila\, P.\, & Giri\, K. (2009). Confidence intervals in regression util
izing prior information. Journal of Statistical Planning and Inference\, 1
39\, 3419-3429.\n Kabaila P. and Ranathunga N. (2019) On Adaptive Gauss
-Hermite Quadrature for Estimation in GLMM’s. In: Nguyen H. (eds) Statis
tics and Data Science. RSSDS 2019. Communications in Computer and Informat
ion Science\, vol 1150. Springer\, Singapore.\n Kabaila\, P.\, & Ranath
unga\, N. (2020). Computation of the expected value of a chi-distributed r
andom variable. Computational Statistics.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/15/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Jason Grealey (La Trobe University\, Baker Institute)
DTSTART;VALUE=DATE-TIME:20210312T010000Z
DTEND;VALUE=DATE-TIME:20210312T020000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/16
DESCRIPTION:Title: Quantifying computational carbon footprints and deep lear
ning in genomic prediction\nby Jason Grealey (La Trobe University\, Ba
ker Institute) as part of La Trobe University Statistics and Stochastic zo
om seminar\n\n\nAbstract\nThis presentation details the three main project
s undertaken within my PhD. The first involves investigating the carbon fo
otprint of computation. As climate change is an extremely pressing global
issue\, researchers must be prudent with energy usage\, this includes comp
utational research. In this first project we developed a freely available
and simple to use carbon footprint estimator of computational tools called
Green Algorithms\, it provides interpretable metrics to understand any gi
ven carbon footprint. The next section I will talk about involves the esti
mation of the carbon footprints of various bioinformatic analyses using pu
blished benchmarks. These carbon footprints are largely unknown and undera
ppreciated within the research community\, we also provide a list of reali
stic and practical recommendations that computational researchers can util
ise in order to minimise their carbon footprint. The last section is a sim
ulation study aiming to understand what types of genetic architectures and
study designs are needed to utilise neural networks in place of tradition
al linear polygenic scoring methods in genomic prediction.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/16/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Nicholas Sing (La Trobe University\, Baker Institute)
DTSTART;VALUE=DATE-TIME:20210401T010000Z
DTEND;VALUE=DATE-TIME:20210401T020000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/17
DESCRIPTION:Title: Mining Lipidomics for Biological Insight\nby Nicholas
Sing (La Trobe University\, Baker Institute) as part of La Trobe Universi
ty Statistics and Stochastic zoom seminar\n\n\nAbstract\nLipidomics is the
study of all lipids that make up cells and organisms. Abnormal lipid meta
bolism is associated with many cardiovascular risk factors. The Baker Hea
rt and Diabetes Institute has generated lipidomic datasets for several pop
ulation studies. These datasets can contain hundreds of lipid species and
sample numbers ranging from hundreds to several thousands. During lipidomi
c analysis unwanted variation can arise due to variation from technical so
urces\, which unwanted variation removal algorithms aim to minimise. This
project aims to develop multivariate methodologies for dealing with unwant
ed variation in lipidomic datasets and modelling the metabolic association
s between groups of lipid species and participant characteristics. We inte
nd to use eigenlipids to explore the existence\, onset or progression of m
etabolic disease. We have demonstrated that eigenlipids can outperform man
y individual lipid species in predicting cardiovascular risk factors. To i
dentify technical sources of unwanted variation in the plasma lipidome dur
ing laboratory processing we recently performed a laboratory experiment\,
which will support the utilisation of unwanted variation removal algorithm
s for removing variation from laboratory processing in pre-existing datase
ts.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/17/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Mitra Jazayeri (La Trobe University)
DTSTART;VALUE=DATE-TIME:20210506T020000Z
DTEND;VALUE=DATE-TIME:20210506T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/18
DESCRIPTION:Title: Factors affecting first year psychology students’ stati
stics learning\nby Mitra Jazayeri (La Trobe University) as part of La
Trobe University Statistics and Stochastic zoom seminar\n\n\nAbstract\nTea
ching statistics to different disciplines is increasingly challenging. Thi
s is due to several factors including the wide range of students’ academ
ic backgrounds\, availability of data and public perception of its importa
nce. In addition\, advancements in technology and recent technological inn
ovations in teaching also present challenges due to the large gap between
learning theory and teaching practices. \nFurthermore\, experiencing anxi
ety when studying statistics\, as a prerequisite subject\, has always been
commonplace for students around the world. Statistics anxiety can appear
as a complex array of emotional reactions from only a minor discomfort to
severe forms of apprehension\, fear\, nervousness\, panic and worry. Cons
idering that statistics is often required as a core subject in a wide rang
e of university degrees\, research into assisting in overcoming these chal
lenges is essential. \nThis research aims to explore intervention methods
to minimize students’ apprehension in their learning process. This is p
resented in three parts: 1) an examination of the effect of blended deliv
ery of an introductory statistics subject\, 2) a systematic review investi
gating interventions utilized to reduce students’ statistics anxiety\, 3
a) the introduction of a survey tool for evaluation of student attitudes\
, confidence\, anxiety\, and beliefs about the usefulness of learning stat
istics in their degree and an assessment of its’ reliability and validi
ty\, 3b) design\, implementation and analysis of a web-based mindfulness i
ntervention delivered to a sample of 530 students studying statistics for
psychology during COVID-19 era. This project will help inform educators
for the better delivery of statistics to students with diverse academic ba
ckgrounds.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/18/
END:VEVENT
BEGIN:VEVENT
SUMMARY:A/Prof Minh-Ngoc Tran (University of Sydney)
DTSTART;VALUE=DATE-TIME:20210617T020000Z
DTEND;VALUE=DATE-TIME:20210617T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/19
DESCRIPTION:Title: Variational Bayes on Manifolds\nby A/Prof Minh-Ngoc T
ran (University of Sydney) as part of La Trobe University Statistics and S
tochastic zoom seminar\n\n\nAbstract\nVariational Bayes (VB) has become a
widely-used tool for Bayesian inference in statistics and machine learning
. Nonetheless\, the development of the existing VB algorithms is so far ge
nerally restricted to the case where the variational parameter space is Eu
clidean\, which hinders the potential broad application of VB methods. Thi
s paper extends the scope of VB to the case where the variational paramete
r space is a Riemannian manifold. We develop an efficient manifold-based V
B algorithm that exploits both the geometric structure of the constraint p
arameter space and the information geometry of the manifold of VB approxim
ating probability distributions. Our algorithm is provably convergent and
achieves a decent convergence rate. We develop in particular several manif
old VB algorithms including Manifold Gaussian VB and Stiefel Neural Networ
k VB\, and demonstrate through numerical experiments that the proposed alg
orithms are stable\, less sensitive to initialization and compares favoura
bly to existing VB methods. This is a joint work with Dang Nguyen and Duy
Nguyen.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/19/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Masters Students (La Trobe University)
DTSTART;VALUE=DATE-TIME:20210624T020000Z
DTEND;VALUE=DATE-TIME:20210624T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/20
DESCRIPTION:Title: Masters Students Talks\nby Masters Students (La Trobe
University) as part of La Trobe University Statistics and Stochastic zoom
seminar\n\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/20/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Dr Matias Quiroz (University of Technology Sydney)
DTSTART;VALUE=DATE-TIME:20210819T020000Z
DTEND;VALUE=DATE-TIME:20210819T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/21
DESCRIPTION:Title: Spectral Subsampling MCMC for Stationary Multivariate Tim
e Series\nby Dr Matias Quiroz (University of Technology Sydney) as par
t of La Trobe University Statistics and Stochastic zoom seminar\n\n\nAbstr
act\nSpectral subsampling MCMC was recently proposed to speed up Markov ch
ain Monte Carlo (MCMC) for long stationary univariate time series by subsa
mpling periodogram observations in the frequency domain. This talk present
s an extension of the approach to stationary multivariate time series. We
also propose a multivariate generalisation of the autoregressive tempered
fractionally differentiated moving average model (ARTFIMA). The new model
is shown to provide a better fit compared to multivariate autoregressive m
oving average models for three real world examples. We demonstrate that sp
ectral subsampling may provide up to two orders of magnitude faster estima
tion\, while retaining MCMC sampling efficiency and accuracy\, compared to
spectral methods using the full dataset.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/21/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Prof Christopher Drovandi (Queensland University of Technology)
DTSTART;VALUE=DATE-TIME:20210909T020000Z
DTEND;VALUE=DATE-TIME:20210909T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/22
DESCRIPTION:Title: Statistical Inference for Implicit Models using Bayesian
Synthetic Likelihood\nby Prof Christopher Drovandi (Queensland Univers
ity of Technology) as part of La Trobe University Statistics and Stochasti
c zoom seminar\n\n\nAbstract\nImplicit models are defined as those that ca
n be simulated but the associated likelihood function is intractable. Suc
h models are prevalent in many fields such as biology\, ecology\, cosmolog
y and epidemiology. Given the unavailability of the likelihood function\,
statistical inference for implicit models is challenging as we must rely
only on the ability to generate mock datasets from the model of interest\,
and compare it with the observed data in some way. This talk will explai
n a useful method called Bayesian synthetic likelihood for conducting such
statistical inference. I will discuss how BSL can be extended to reduce
the number of model simulations required and to make it more robust to mod
el misspecification. I will also describe some theoretical properties of
the method.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/22/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Illia Donhauzer (La Trobe University)
DTSTART;VALUE=DATE-TIME:20210916T020000Z
DTEND;VALUE=DATE-TIME:20210916T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/23
DESCRIPTION:Title: On the asymptotic behavior of functionals of random field
s\nby Illia Donhauzer (La Trobe University) as part of La Trobe Univer
sity Statistics and Stochastic zoom seminar\n\n\nAbstract\nThe talk is abo
ut the asymptotic behavior of functionals of random fields with possible l
ong-range dependence. New properties of generalized Hermite-type processes
\, the Strong Law of Large Numbers (SLLN) for random fields\, and the asym
ptotic behavior of running maxima of random double arrays will be discusse
d.\n\nNew properties of generalized Hermite-type processes that arise in N
LT for integral functionals of long-range dependent random\n\nfields will
be demonstrated. Contrary to the classical one-dimensional case\, it will
be shown that for any choice of a multidimensional observation window the
generalized Hermite-type process has non-stationary increments.\n\nThe SLL
N for integral functionals of random fields with unboundedly increasing co
variances will be presented. The SLLN is derived for the case of increasin
g domains. Conditions on covariance functions such that the SLLN holds wil
l be provided. The considered scenarios include non-stationary random fiel
ds. The discussion about applications to weak and long-range dependent ran
dom fields and numerical examples will be shown.\n\nResults on the asympto
tic behavior of running maxima functionals of random double arrays of phi-
subgaussian random variables will be demonstrated. The main results are sp
ecified for various important particular scenarios and classes of phi-subg
aussian random variables.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/23/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Dr. Miryana Grigorova (University of Leeds)
DTSTART;VALUE=DATE-TIME:20210930T090000Z
DTEND;VALUE=DATE-TIME:20210930T100000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/24
DESCRIPTION:Title: Superhedging of options in a non-linear incomplete financ
ial market model.\nby Dr. Miryana Grigorova (University of Leeds) as p
art of La Trobe University Statistics and Stochastic zoom seminar\n\n\nAbs
tract\nWe will study the superhedging price (and superhedging strategies)
of European and American options in a non-linear incomplete market model w
ith default\, with a particular focus on the American options case which i
s more involved. We will provide a dual representation of the seller’s
(superhedging) price for the American option in terms of a mixed stochasti
c control/stopping problem with non-linear expectations/ evaluations\, and
in terms of non-linear Reflected BSDEs with constraints. If time permits\
, we will also present a duality result for the buyer’s price in terms o
f a stochastic game of control and stopping with non-linear expectations/
evaluations.\n\nZoom meeting link:\n\nhttps://unimelb.zoom.us/j/8695143126
9?pwd=S1FPSFBHLzd5QkpGYlJIYS9wUGtLUT09\n\n(if the link doesn't work when y
ou click it -- please copy & paste it into the address bar in your browser
).\n\nPassword: 422668 (just in case)\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/24/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Dr. Francis K.C. Hui (Australian National University)
DTSTART;VALUE=DATE-TIME:20211021T010000Z
DTEND;VALUE=DATE-TIME:20211021T020000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/25
DESCRIPTION:Title: Spatio-temporal joint species distribution modeling – A
community-level basis function approach\nby Dr. Francis K.C. Hui (Aus
tralian National University) as part of La Trobe University Statistics and
Stochastic zoom seminar\n\n\nAbstract\nThe last decade in ecology has see
n the development and rising popularity of joint species distribution mode
ling\napproaches for studying species assemblages\, with by far the most c
ommon approach being based around\ngeneralized linear latent variable mode
ls (LVMs). However\, while methodological and computational advances\ncont
inue to be made with LVMs\, their application to spatio-temporal multivari
ate abundance data i.e.\, observations\nof multiple species recorded acros
s space and/or time\, remains computationally challenging and not necessar
ily\nscalable when it comes to fitting and inference.\n\nIn this talk\, we
propose an alternative approach to spatio-temporal joint species distribu
tion modeling which breaks\naway from the LVM framework. Inspired by the c
oncept of fixed rank kriging\, we employ a set of fixed\, communitylevel\n
spatial and/or temporal basis functions\, with corresponding species-speci
fic random slopes to account for\nspatio-temporal correlations both within
and between species. The resulting community-level basis function model\n
(CBFM) can be used for the same array of purposes as LVMs\, but is designe
d to be computationally much more\nefficient given they can be set up and
thus fitted using the same machinery as for generalized additive models.\n
Simulations and an application to a demersals fish dataset collected off t
he Northeast US continental shelf illustrate\nthe potential of CBFMs for s
calable spatio-temporal joint species distribution modeling.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/25/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Masters Students (La Trobe University)
DTSTART;VALUE=DATE-TIME:20211104T010000Z
DTEND;VALUE=DATE-TIME:20211104T020000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/26
DESCRIPTION:Title: Masters Students Talks\nby Masters Students (La Trobe
University) as part of La Trobe University Statistics and Stochastic zoom
seminar\n\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/26/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Mitra Jazayeri (La Trobe University)
DTSTART;VALUE=DATE-TIME:20211216T040000Z
DTEND;VALUE=DATE-TIME:20211216T050000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/27
DESCRIPTION:Title: Validity analysis of a modified extended Technology Accep
tance Model for first year Psychology students\nby Mitra Jazayeri (La
Trobe University) as part of La Trobe University Statistics and Stochastic
zoom seminar\n\n\nAbstract\nThis talk continues on from my confirmation w
here an outline of my research and the results for the first two phases of
my project were given. In this talk I predominantly present the steps tak
en to evaluate the survey tool\, which is a modified and extended Technolo
gy Acceptance Model (TAM). This measurement scale determines the perceptio
n of psychology students about the ease of use and usefulness of statistic
al concepts and their application in psychology using the statistical soft
ware\, SPSS.\n\nThe proposed model was tested for its reliability and stru
ctural validity using data collected from a survey of first year psycholog
y students studying statistics during the global pandemic in 2020. To expl
ore the structure of the constructs of students’ attitude\, confidence a
nd perception\, an Exploratory Factor Analysis (EFA) was conducted on the
responses data set. Five latent variables were identified. Utilizing maxim
um likelihood estimates in Confirmatory Factor Analysis (CFA)\, and Analys
is of Moment Structures (AMOS)\, the results supported the proposed EFA mo
del. In addition\, results of the CFA indicated that the best fitted model
had correlations among four of the five constructs. Internal consistency
estimates utilizing alpha coefficients\, ranged from 0.81 to 0.88 with onl
y one exception of 0.682. The findings provide a valid and reliable assess
ment of students’ attitudes towards statistics for predicting academic p
erformance. Consequently\, this may help as a guide for effective decision
-making in the design and development of the statistics subjects for stude
nts with a non-mathematical background.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/27/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Yeshna Bhowon (La Trobe Universty)
DTSTART;VALUE=DATE-TIME:20220223T223000Z
DTEND;VALUE=DATE-TIME:20220223T233000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/28
DESCRIPTION:Title: Applications Of Data Science Methods Within A Community-B
ased Physical Activity Program For Young People With Disability\nby Ye
shna Bhowon (La Trobe Universty) as part of La Trobe University Statistics
and Stochastic zoom seminar\n\n\nAbstract\nFitSkills is a community-based
program that connects university student mentors to young people living w
ith disability through exercise programs at their local community gyms. Ac
cess to exercise facilities is a commonly documented perceived barrier to
participation in physical activity for people living with disability\, but
the problem has not been quantified. We conducted a geospatial analysis u
sing a population cohort in an aim to quantify this perceived barrier. The
second part of my research used data collected during the FitSkills trial
to determine if completing FitSkills fostered positive attitudes towards
disability among the student mentors.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/28/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Nicholas Sing (La Trobe University and Baker Heart and Diabetes In
stitute)
DTSTART;VALUE=DATE-TIME:20220407T020000Z
DTEND;VALUE=DATE-TIME:20220407T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/29
DESCRIPTION:Title: Mining Lipidomics for Biological Insight\nby Nicholas
Sing (La Trobe University and Baker Heart and Diabetes Institute) as part
of La Trobe University Statistics and Stochastic zoom seminar\n\n\nAbstra
ct\nLipidomics is the study of all lipids that make up cells and organisms
. Abnormal lipid metabolism is associated with many cardiovascular risk fa
ctors. The Baker Heart and Diabetes Institute has generated lipidomic data
sets for several population studies. During lipidomic analysis unwanted va
riation can arise due to variation in laboratory processing and handling\,
which unwanted variation removal algorithms aim to minimise. This project
aims to develop multivariate methodologies for dealing with unwanted vari
ation in lipidomic datasets and modelling the metabolic associations betwe
en groups of lipid species and participant characteristics. In this projec
t we are using lipid set enrichment analysis and eigenlipids to explore li
pid biology associated with cardiovascular disease. To identify technical
sources of unwanted variation in the plasma lipidome during laboratory pro
cessing we performed a laboratory experiment and utilised the Remove Unwan
ted Variation-III (RUV-III) algorithm to remove these sources of unwanted
variation from the lipidomic dataset we acquired. We intend to use this as
a basis to identify negative control lipids to remove similar sources of
unwanted variation in population lipidomic datasets using RUV-III.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/29/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Dr Alan Huang (University of Queensland)
DTSTART;VALUE=DATE-TIME:20220428T020000Z
DTEND;VALUE=DATE-TIME:20220428T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/30
DESCRIPTION:Title: On arbitrarily underdispersed discrete distributions\
nby Dr Alan Huang (University of Queensland) as part of La Trobe Universit
y Statistics and Stochastic zoom seminar\n\n\nAbstract\nWe review a range
of generalized count distributions\, investigating which (if any) can be a
rbitrarily underdispersed\, i.e.\, its variance can be arbitrarily small c
ompared to its mean. A philosophical implication is that models failing th
is criterion perhaps should not be considered a “statistical model” ac
cording to the extendibility criterion of McCullagh (2002). Four practical
implications will be discussed. We suggest that all generalizations of th
e Poisson distribution be tested against this property.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/30/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Ayesha Perera (La Trobe Univerity)
DTSTART;VALUE=DATE-TIME:20220616T020000Z
DTEND;VALUE=DATE-TIME:20220616T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/31
DESCRIPTION:Title: Performance of Model Averaged Tail Area Confidence Interv
al\nby Ayesha Perera (La Trobe Univerity) as part of La Trobe Universi
ty Statistics and Stochastic zoom seminar\n\n\nAbstract\nEvery model has a
n uncertainty in the variables that it should include. Model averaging is
considered as a promising method that could be used to perform inference i
n the presence of model uncertainty. The performance of this method heavil
y depends on the data-based model weights used. Traditionally\, this weigh
t is chosen to be proportional to the exponential of minus the Generalized
Information Criterion divided by two. We observe that the model-based con
fidence interval performs better\, in terms of coverage and expected lengt
h\, in the case of two nested linear regression models when this division
by two is replaced by multiplied by a positive tuning constant. In the sec
ond part of the talk\, we extend the analysis of the performance of Model
Averaged Tail Area confidence interval by Kabaila\, Welsh and Abeysekara\,
Scandinavian Journal of Statistics\, 2016\, to the case of three or more
nested linear regression models. We also assess the influence of the weigh
t function on the performance of this confidence interval for three nested
linear regression models applied to the ‘Cholesterol’ data set.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/31/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Masters students talks (La Trobe Univerity)
DTSTART;VALUE=DATE-TIME:20220623T020000Z
DTEND;VALUE=DATE-TIME:20220623T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/32
DESCRIPTION:Title: 2 masters theses presentation\nby Masters students ta
lks (La Trobe Univerity) as part of La Trobe University Statistics and Sto
chastic zoom seminar\n\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/32/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Faraz Fattah Hesari (La Trobe Univerity)
DTSTART;VALUE=DATE-TIME:20220630T020000Z
DTEND;VALUE=DATE-TIME:20220630T030000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/33
DESCRIPTION:Title: Statistical modelling of property prices\, applying genet
ic algorithm and isolation forests\nby Faraz Fattah Hesari (La Trobe U
niverity) as part of La Trobe University Statistics and Stochastic zoom se
minar\n\nAbstract: TBA\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/33/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Dr Boris Buchmann (Australian National University)
DTSTART;VALUE=DATE-TIME:20220804T070000Z
DTEND;VALUE=DATE-TIME:20220804T080000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/34
DESCRIPTION:Title: Weak Subordination of Multivariate Levy Processes\nby
Dr Boris Buchmann (Australian National University) as part of La Trobe Un
iversity Statistics and Stochastic zoom seminar\n\n\nAbstract\nYou are wel
come to attend the following Statistics and Stochastic colloquium (part of
the Colloquium Series of the Department of Mathematics and Statistics) at
La Trobe Universit\, which is jointly organized with the Probability Vict
oria Seminar.\n\nPVSeminar #36\, Thursday 04 August / 17:00 AEST \n\nBor
is Buchmann (Australian National University\, Australia): Weak Subordinati
on of Multivariate Levy Processes \n\nAbstract: Subordination is the opera
tion which evaluates a Levy process at a subordinator\, giving rise to a p
athwise construction of a "time-changed" process. Originating with Bochner
in the context of probability semigroups\, subordination was applied by M
adan and Seneta to create the variance gamma process\, which is prominentl
y used in financial modelling. However\, unless the subordinate has indepe
ndent components or the subordinator has indistinguishable components\, su
bordination may not produce a Levy process. \n\nWe introduce a new operat
ion known as weak subordination that always produces a Levy process by ass
igning the distribution of the subordinate conditional on the value of the
subordinator\, and matches traditional subordination in law in the cases
above. Weak subordination is applied to extend the class of variance gener
alised gamma convolutions and to construct the weak variance-alpha-gamma p
rocess. The latter process exhibits a wider range of dependence than using
traditional subordination. \n\nJoint work with Kevin W Lu (UW)\, Dilip B
Madan (UM)\, Marcus Michaelsen (UHH)\, Adam Nie (NTU)\, Alex Szimayer (UH
H). \n\nZoom meeting link: https://unimelb.zoom.us/j/83757047993?pwd=a04z
NitYZTRHdTZYdERkMmJYdDRWZz09\n \n(if the link do
esn't work when you click it -- please copy & paste it into the address ba
r in your browser).\n\nPassword: 916563 (just in case)\n\nA PDF file w
ith the talk slides might become available for downloading from our semina
r Webpage at https://probvic.wordpress.com/pvseminar/ prior to the talk (t
he above Zoom is being posted there).\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/34/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Prof Yimin Xiao (Michigan State University\, USA)
DTSTART;VALUE=DATE-TIME:20220818T000000Z
DTEND;VALUE=DATE-TIME:20220818T010000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/35
DESCRIPTION:Title: Sample Path and Extreme Value Properties of Multivariate
Gaussian Random Fields\nby Prof Yimin Xiao (Michigan State University\
, USA) as part of La Trobe University Statistics and Stochastic zoom semin
ar\n\n\nAbstract\nIn this talk\, we present some recent results on sample
path and extreme value properties of a large class of multivariate Gaussia
n random fields including multivariate Gaussian fields\, operator fraction
al Brownian motion\, vector-valued operator-scaling random fields\, and ma
trix-valued Gaussian random fields. These results illustrate explicitly th
e effects of the dependence structures among the coordinate processes on t
he sample path and extreme value properties of multivariate Gaussian rando
m fields.\n\nZoom meeting link: https://unimelb.zoom.us/j/86460269383?pwd=
aDNWbk4yWDdzclhUOWZ6ZElFQnlrQT09 \n \n(if the ab
ove link doesn't work when you click it -- please copy & paste it into the
address bar in your browser).\n\nPassword: 457925 (just in case)\n\nA PDF
file with the talk slides might become available for downloading from our
seminar Webpage at https://probvic.wordpress.com/pvseminar/ prior to the
talk (the above Zoom link has already been posted there).\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/35/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Prof Giovanni Peccati (University of Luxembourg)
DTSTART;VALUE=DATE-TIME:20220908T070000Z
DTEND;VALUE=DATE-TIME:20220908T080000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/36
DESCRIPTION:Title: Some variations on a theme by P. de Jong\nby Prof Gio
vanni Peccati (University of Luxembourg) as part of La Trobe University St
atistics and Stochastic zoom seminar\n\n\nAbstract\nJoint La Trobe statist
ics and stochastics and PV seminar\n\nAbstract: In a remarkable paper from
1990\, the Dutch mathematician P. de Jong proved a striking Central Limit
Theorem yielding that\, for a sequence of normalized and degenerate U-sta
tistics verifying a Lindeberg-type condition\, convergence to Gaussian is
equivalent to the convergence of their fourth cumulants to zero. Such a re
sult is the ancestor of the collection of “fourth-moment theorems” for
non-linear functionals of random fields\, that have recently played a pro
minent role in several questions of mathematical physics and stochastic ge
ometry. In my talk\, I will first present some quantitative multidimension
al extensions of de Jong’s result\, obtained by using Stein’s method o
f exchangeable pairs. I will then discuss some recent functional versions
of de Jong’s findings\, both in the symmetric and non-symmetric cases. T
he results in the symmetric case yield some novel universality results for
U-processes\, generalizing a classic invariance principle by Miller and S
en (1972)\, and allowing one to establish a complete taxonomy of functiona
l CLTs associated with counting statistics of random geometric graphs. My
presentation is mainly based on the following references:\n\nCh. Döbler a
nd G. Peccati: Quantitative de Jong Theorems in any dimension. EJP\, 2016.
\n\nCh. Döbler\, M. Kasprzak and G. Peccati: Weak convergence of U-proces
ses with size-dependent kernels. Ann. App. Prob.\, 2022\n\nCh. Döbler\, M
. Kasprzak and G. Peccati. The multivariate functional de Jong CLT. Probab
. Th. Rel. Fields\, 2022+\n\nZoom meeting link: https://unimelb.zoom.us/j/
82317899187?pwd=TThhQmZrcGtxSGpQL2wzTHJjZlZjQT09\n
\n(if the above link doesn't work when you click it -- please copy & pa
ste it into the address bar in your browser).\n\nPassword: 633070 (just in
case)\n\nA PDF file with the talk slides might become available for downl
oading from our seminar Webpage at https://probvic.wordpress.com/pvseminar
/ prior to the talk (the above Zoom link will also be posted there shortly
).\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/36/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Honours and Masters students (La Trobe University)
DTSTART;VALUE=DATE-TIME:20221027T010000Z
DTEND;VALUE=DATE-TIME:20221027T023000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/37
DESCRIPTION:Title: Honours and Masters theses presentations\nby Honours
and Masters students (La Trobe University) as part of La Trobe University
Statistics and Stochastic zoom seminar\n\n\nAbstract\nThe schedule is the
following (20min for Thesis B talks and 10min for Thesis A talk):\n\n\n12.
05pm Adam Bilchouris. Investigating Statistical Properties of Functionals
of Strongly Dependent Spatial Data.\n\n12.15pm Lennon Zachary Logan. Mathe
matics of Kirigami. A Study of Euclidean Nets and Hyperbolic Crystals.\n\n
12.35pm Dmytro Ostapenko. Statistical Modelling of ANZ Property Data.\n\n1
2.55pm Juliet Nwabuzor. Performance of Preliminary Model Selection Using B
ayesian Information Criterion (Bic).\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/37/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Mitra Jazayeri (La Trobe university)
DTSTART;VALUE=DATE-TIME:20221214T233000Z
DTEND;VALUE=DATE-TIME:20221215T003000Z
DTSTAMP;VALUE=DATE-TIME:20230208T080017Z
UID:StatisticsandStochastic/38
DESCRIPTION:Title: Teaching statistics in a new era- factors affecting stude
nts’ learning and aligning technology with how we teach\nby Mitra Ja
zayeri (La Trobe university) as part of La Trobe University Statistics and
Stochastic zoom seminar\n\n\nAbstract\nThis thesis investigates the facto
rs that affect psychology students’ ability to learn statistics in a new
era of technological advancements and the effects post-Covid on education
. Historically\, teaching statistics to psychology students has been one o
f the most challenging tasks for statistics educators worldwide. Previous
research has been conducted around the theme of social science students’
statistics anxiety and the varied survey designs developed to measure it.
However\, little has been done to reduce statistics anxiety\, with the ai
m of increasing the performance of students with a non-mathematical backgr
ound\, particularly in this technological age.\n\n \nThe aim of this rese
arch is to i) conduct multiple regression and sub-group analyses using the
R software package to investigate whether the blended delivery of a 12-we
ek statistics subject to first-year psychology students had any effect on
performance compared to face-to-face teaching only\; ii) design a mindfuln
ess intervention\, together with a step-by-step methodological approach fo
r teaching statistics to first-year psychology students\; iii) develop a
survey based on the technology acceptance model to measure students’ anx
iety which included testing the validity and reliability of the adopted su
rvey tool. To do so\, the structural properties of the survey were investi
gated. For this stage of the research\, jamovi and the IBM SPSS AMOS softw
are package were utilized to obtain Cronbach’s alpha and the exploratory
and confirmatory factor analysis output. This thesis finds that the web-b
ased mindfulness intervention had a significant positive effect on student
s’ statistics anxiety and therefore performance. Moreover\, five constru
cts were identified which affect students’ statistics anxiety and theref
ore their performance\, namely attitude\, confidence\, student’s awarene
ss of their mental state\, independent learner belief\, and dependent lear
ner belief. The findings of this research may assist and inspire statistic
s educators internationally in their approach to the design and developmen
t of their teaching material to non-mathematical students for whom statist
ics is a core subject in their study.\n
LOCATION:https://researchseminars.org/talk/StatisticsandStochastic/38/
END:VEVENT
END:VCALENDAR