Speaker Abstracts
Yan Bai, Ph.D. Graduate, Department of
Statistics
Region Adaptive algorithm with Online Recursion
The efficiency of Markov chain Monte Carlo (MCMC) algorithms can
vary dramatically with the choice of simulation parameters. Adaptive
MCMC (AMCMC) algorithms allow the automatic tuning of the parameters
while the simulation is in progress. In the case of the target with
multimodes, we propose an adaptation process which involves fitting
the mixture using the available samples via an online EM algorithm
and, based on the current mixture parameters. The method is compared
with other regional AMCMC samplers and is tested on simulated as
well as some examples.
____________________________
Billy Chang, Ph.D. Candidate, Department
of Public Health Sciences
Statistical Approach to Transcript Quantification in RNA-Seq
Recently developed Next Generation Sequencing techniques, such
as the RNA-Seq protocol, provide unprecedented precision and throughput
for transcriptome analysis. The data generative mechanism implied
by RNA-Seq however differs substantially from its predecessors,
such as microarrays. This calls for new methodological development
for analyzing data created under the new sequencing protocol. In
this talk I will focus on the problem of transcript quantification,
i.e. how to quantify the amount of transcripts within a sample of
DNA from a molecular/cell content. After a brief introduction to
the RNA-Seq sequencing pipeline, I will discuss some of the challenges
involved in transcript quantification under the RNA-Seq protocol,
and how statistical methods may resolve the problems at hand.
____________________________
Madeleine Thompson, Ph.D. Candidate,
Department of Statistics
Using Correlation Length to Compare MCMC Methods Graphically
Comparisons of Markov Chain Monte Carlo methods that use tables
of figures of merit are limited by the number of simulation results
that can be displayed without overwhelming the reader. I will demonstrate
a method for comparing MCMC methods that uses grids of plots of
estimated correlation length to communicate the results of a diverse
collection of simulations in a way that readers can interpret easily,
enabling them to more easily identify an MCMC method suitable for
a given task.
____________________________
Angel Valov, PhD Graduate, Department
of Statistics, University of Toronto
First Passage Times for Brownian Motion with Applications
The first time a continuous stochastic process crosses a given,
time-dependent, boundary function is known as a first-passage time
(FPT) for the process. FPTs give rise to stopping time problems
which have a long history and which have recently received renewed
attention from both industry and academia due to their applicability
in finance. In this talk I will briefly outline the classical FPT
problems for Brownian motion as well as discuss a more recent variation
of the original problems. In addition, I will present some 'real
world' applications as well as sketch a number of potential applications
from finance and statistics.
