Statistical Laboratory Seminars

Michaelmas Term 2006

Centre for Mathematical Sciences
Wilberforce Road, Cambridge, CB3 0WB
Tel: (01223) 337958
Fax: (01223) 337956
Email: secretary [at] statslab.cam.ac.uk

Seminars will be held in Meeting Room 12

All interested are welcome


Seminar Schedule:

This list is subject to revision

Select a date to view the relevant seminar abstracts:

Tuesday 10 October
1.00 pm Mike West (Duke)
Sparse statistical modelling and stochastic computation in `large p' problems
Friday 13 October
2.00 pm Anthony Davison (Lausanne)
Smoothing and temperature extremes
Tuesday 17 October
2.00 pm Phil Dawid (UCL)
Statistical causality: conceptions, connexions, confusions, contentions
Friday 27 October
2.00pm Zoubin Ghahramani (Department of Engineering, University of Cambridge)
Bayesian nonparametric latent feature models
Friday 3 November
2.00 pm Darren Upton and Steffan Berridge, AHL Research, Man Investments Ltd
Examples of quantitative research in the hedge fund industry
Friday 10 November
2.00 pm Aurore Delaigle (Bristol)
Nonparametric regression with coarsened predictors
Please note this seminar will be held in MR3 NOT MR12
Friday 17 November
2.00 pm Idris Eckley and Yanyun Wu (Shell)
Neither finance nor pharma - life as an industrial statistician

Abstracts and Further Details:

Tuesday 10 October

1.00pm Mike West (Duke)

Sparse statistical modelling and stochstic computation in `large p' problems
The concepts and methods of Bayesian sparsity modelling are key to problems of variable selection and multivariate structure assessment in statistical science, and provide foundation for scaling statistical models from the viewpoints of both scientific parsimony and computational accessibility/feasibility. The ability to induce appropriate degrees and relevant structure of sparsity in multivariate modelling is a critical concept if statistical methodology is to scale with increasing dimension. My talk will overview a range of developments in the application of sparsity modelling in three key, related areas of modern multivariate analysis: (highly) multivariate regression, high-dimensional latent factor analysis, and graphical modelling. I will review and develop modelling and computational questions, the latter involving stochastic search and MCMC for "big" and "sparse" models, and draw on motivating applications in sparsity modelling in problems from genomics, involving biological pathway studies, and multivariate time series analysis.

Friday 13 October

2.00pm Anthony Davison (Lausanne)

Smoothing and temperature extremes
Smoothing methods are now well-established in many domains of statistics, and are increasingly used in analysis of extremal data. The talk will describe some applications of smoothing to data on temperature extremes, elucidating the relation between cold winter weather in the Alps and the North Atlantic Oscillation, and changes in the lengths of unusually hot and cold spells in central England. The work mixes classical models for extremes, generalised additive modelling, local polynomial smoothing, and the bootstrap, and is joint with Valérie Chavez-Demoulin and Mariá Süveges.

Tuesday 17 October

2.00pm Phil Dawid (UCL)

Statistical causality: conceptions, connexions, confusions, contentions
Modern statistical approaches to causal inference are based on a variety of distinct foundations, ingredients, assumptions and methods. These involve differing conceptions of the effects of interventions, or of stable relationships across regimes; disagreement over the roles of hypothetical and counterfactual outcomes; and varying semantics and uses for algebraic, graphical and other representations. There does however seem to be fairly broad agreement that causal inference requires significant modifications and extensions to standard statistical machinery. I shall argue that this is mistaken, and that the power of existing statistical and decision-theoretic tools to address causal issues is much greater than is commonly allowed.

Friday 27 October

2.00pm Zoubin Ghahramani (Department of Engineering, University of Cambridge)

Bayesian nonparametric latent feature models
Latent variables are an important component of many statistical models. Most latent variable models, such as mixture models, factor analysis, and independent components analysis, assume a finite, usually small number of latent variables. However, it may be statistically undesirable to constrain the number of latent variables a priori. Here we show how a more flexible nonparametric approach is possible in which the number of latent variables is unbounded. To do this, we describe a probability distribution over equivalence classes of binary matrices with a finite number of rows, corresponding to the data points, and an unbounded number of columns, corresponding to the latent variables. Each data point can be associated with a subset of the possible latent variables, which we refer to as the latent features of that data point. The binary variables in the matrix indicate which latent feature is possessed by which data point, and there is a potentially infinite array of features. We derive the distribution over unbounded binary matrices by taking the limit of a distribution over N × K binary matrices as K → ∞, a strategy inspired by the derivation of the Chinese restaurant process (Aldous, 1985; Pitman, 2002) which preserves exchangeability of the rows. We define a simple generative processes for this distribution which we call the Indian buffet process (IBP; Griffiths and Ghahramani, 2005). We describe recent extensions of this model, Markov chain Monte Carlo algorithms for inference, and a number of applications to collaborative filtering, bioinformatics, cognitive modelling, and causal discovery.

Joint work with Thomas L. Griffiths, UC Berkeley.

Friday 3 November

2.00pm Darren Upton and Steffan Berridge, AHL Research, Man Investments Ltd.

Examples of quantitative research in the hedge fund industry
Darren will begin by giving a brief overview of the hedge fund industry and the different styles of investment this encompasses. He will then illustrate some of the opportunities that exist for quantitative research in the industry using examples of projects completed within the AHL research team. One particular research project involves designing a strategy to trade the volatility of currencies, and Steffan will discuss how we might measure the risk of such a strategy. This is not straightforward, since firstly option payoffs are nonlinear functions of the underlying market variables; and secondly there are additional stochastic factors that can greatly affect the value of an option. These added complications mean that analytical methods for risk measurement usually become intractable, and thus Monte Carlo-based measurement of risk is the most attractive.

Friday 10 November

2.00pm Aurore Delaigle (Bristol)

Nonparametric regression with coarsened predictors
We consider nonparametric estimation of a regression function in an errors-in-variables problem. We first review the classical errors-in-variables problem, where the predictor variable is observed with error, and then make the distinction with our particular problem. There, we are are interested in the reversed problem: we observe a variable accurately, but we are interested in a contaminated version of this variable This is joint work with Peter Hall and Hans Muller.
Please note this seminar will be held in MR3 NOT MR12

Friday 17 November

2.00pm Idris Eckley and Yanyun Wu (Shell)

Neither finance nor pharma - life as an industrial statistician. To be followed by an introduction and invitation to join the Royal Statistical Society

The role of a statistician within a large multinational is very diverse, providing a rich opportunity to work with interesting people in different environments. It also provides many different challenges (both technical and organisational). This talk aims to give you flavour of what life as an industrial statistician is like - from first weeks to possible career paths & challenges.

At the end of the talk there will be a ten minute presentation on the aims and activities of the Royal Statistical Society after which there will be the oppotunity to become a member of the Society.


Seminar organizer, Susan Pitts.
Please see also the Informal Probability Seminars and Network Seminars

Last updated 9-Sept-2005
Valid HTML (4.01 Transitional)