Departmental Colloquium - Statistics, 2009

Thursday 3:30 PM, Chase 319

Chair: Hong Gu

Sep. 17^th,

John Robinson, School of Mathematics and Statistics, University of Sydney.

Title: Asymptotic Approximations in Sampling and Resampling

Abstract: The ideas of sampling with and without replacement connect each of the areas of finite population samples, permutation tests, rank tests and bootstrap methods. Normal and Edgeworth approximations, Cram\' er large deviation results and related saddlepoint approximations have been obtained in one and two sample cases, and some results on Studentized statistics, $k$-sample and multivariate cases have been obtained quite recently. In the case of the bootstrap, asymptotic methods have been used to give theoretical results on their accuracy. I will survey of some of the main asymptotic methods and indicate in more detail some relatively recent results, on approximations for Studentized and multivariate statistics.

 
Oct. 15th,
 
Chris Field, Dalhousie University, Halifax, Canada
 
Title: Bootstrapping Robust Hierarchical Models
 
Abstract: This talk is based on joint research with Alan Welsh and Zhen Pang. In mixed models, the use of robust estimates raises several 
interesting inferential challenges. One of these arises from the realization that the effect of contamination is to increase the variability in 
the data but robust estimates of variance components are usually smaller than their non-robust counterparts. The robust estimates reflect the 
variability of the uncontaminated core of the model which is not the same as the variability in the data generating process. This means that 
the naive implementation of bootstrap procedures may not work. In this paper, we consider several bootstrap procedures including random 
effect, transformation and weighted bootstraps. Conditions for the asymptotic validity of the bootstraps are given and their performance is 
assessed via a small simulation study. Both the transformation and weighted bootstrap perform well and are asymptotically valid under 
reasonable conditions.
 
 
Nov. 26th,
 
 
Yonggan Zhao, Dalhousie University, Halifax, Canada

Title: Predicting Stock Returns with Regime-Switching Risk Factors.

Abstract: We develop a predicative model of stock returns in the presence of regime-switching risk factors. Based on the Bayes Information

Criterion, we first identify the dynamics of the market regimes with the observed macro and micro economic factors.  We then investigate the

relations between stock returns and the fitted values of the risk factors. Finally, we present an empirical analysis on how this model can be applied

in practice.