The Royal Statistical SocietyThe Royal Statistical Society
Preprints of journal discussion papers

The preprints here are available to download to encourage discussion at Ordinary Meetings of the Society before publication in the Journal of the Royal Statistical Society. Certain other papers which will appear in the journal, such as Presidential addresses, are provided for early dissemination of material that may be of wide interest. However, these preprints will typically be removed from this site as soon as they have appeared in print. All preprints available here are provisional and therefore subject to later amendment by the authors. If you would like to receive a preprint for each meeting automatically by e-mail, please contact Abdel Khairoun at journal@rss.org.uk.
 
The meetings are usually held at the Society's premises in London and normally begin at 5.00 p.m., with tea available from 4.30 p.m. Fellows and guests are welcome to attend the free informal drinks reception which follows the Ordinary Meeting at about 7 p.m. The meetings may be preceded by an informal session at 3.00 p.m. on the issues raised by the papers.
 
2010
 
Presidential addresses

Discussion at Ordinary Meetings

Contributions to the discussion at Ordinary Meetings are welcome, whether in person at the meeting or subsequently in writing. If you would like to speak at a meeting, please contact Abdel Khairoun at journal@rss.org.uk, preferably at least a week before the date of the meeting. Contributions must not exceed 5 minutes' speaking time and 400 words for publication in the journal (excluding details of any references quoted). In either case, written versions should be sent to the Executive Editor at the Royal Statistical Society, 12 Errol Street, London, EC1Y 8LX, UK, or by e-mail as PostScript or PDF file attachments to journal@rss.org.uk to arrive no later than 2 weeks after the meeting. If time allows, contributions that are received before the day of the meeting may be read out by the Secretary for the meeting on behalf of anyone who cannot attend.
 
 
Preprints
 
 
N. Meinshausen (University of Oxford) and P. Bühlmann (Eidgenössiche Technische Hochschule Zürich)
 
Stability selection 
 
Estimation of structure, such as in variable selection, graphical modelling or cluster analysis, is notoriously difficult, especially for high dimensional data. We introduce stability selection. It is based on subsampling in combination with (high dimensional) selection algorithms. As such, the method is extremely general and has a very wide range of applicability. Stability selection provides finite sample control for some error rates of false discoveries and hence a transparent principle to choose a proper amount of regularisation for structure estimation. Variable selection and structure estimation improve markedly for a range of selection methods if stability selection is applied. We prove for the randomized lasso that stability selection will be variable selection consistent even if the necessary conditions needed for consistency of the original lasso method are violated. We demonstrate stability selection for variable selection and Gaussian graphical modelling, using real and simulated data.
 
To be published in Series B.
 
Download:
 
 
RESEARCH SECTION ORDINARY MEETING, Wednesday, May 12th, 2010
 
M. Cule and R. Samworth (University of Cambridge) and M. Stewart (University of Sydney)
 
Maximum likelihood estimation of a multi-dimensional log-concave density

Density estimation is fundamental to visualising structure in multivariate data, and has many other applications. We introduce a non-parametric method that, unlike alternatives, is fully automatic, with no smoothing parameters to choose. By imposing the qualitative shape constraint of log-concavity, we obtain an estimate with attractive properties and extensions.
 
To be published in Series  B.
 
Download:
 
 
 
RESEARCH SECTION ORDINARY MEETING, Wednesday, October 13th, 2010
 
M. Girolami (University College London) and B. Calderhead (University of Glasgow)

Riemann manifold Langevin and Hamiltonian Monte Carlo methods

The paper proposes Metropolis adjusted Langevin and Hamiltonian Monte Carlo sampling methods defined on the Riemann manifold to resolve the shortcomings of existing Monte Carlo algorithms when sampling from target densities that may be high dimensional and exhibit strong correlations. The methods provide fully automated adaptation mechanisms that circumvent the costly pilot runs that are required to tune proposal densities for Metropolis-Hastings or indeed Hamiltonian Monte Carlo and Metropolis adjusted Langevin algorithms. This allows for highly efficient sampling even in very high dimensions where different scalings may be required for the transient and stationary phases of the Markov chain. The methodology proposed exploits the Riemann geometry of the parameter space of statistical models and thus automatically adapts to the local structure when simulating paths across this manifold, providing highly efficient convergence and exploration of the target density. The performance of these Riemann manifold Monte Carlo methods is rigorously assessed by performing inference on logistic regression models, log-Gaussian Cox point processes, stochastic volatility models and Bayesian estimation of dynamic systems described by non-linear differential equations. Substantial improvements in the time-normalized effective sample size are reported when compared with alternative sampling approaches. MATLAB code that is available from the authors  allows replication of all the results reported.
 
 
To be published in Series B.
 
Download:
 
 
ORDINARY MEETING, Wednesday, October 20th, 2010
(Note the early start time of 3.00 p.m.)
 
C. J. Wild, M. Pfannkuch and M. Regan (University of Auckland) and N. J. Horton (Smith College, Northampton)

Towards more accessible conceptions of statistical inference

There is a compelling case, based on research in statistics education, for first courses in statistical inference to be underpinned by a staged development path. Preferably over a number of years, students should begin working with precursor forms of statistical inference, much earlier than they now do. A side benefit is giving younger students more straightforward and more satisfying ways of answering interesting real world questions. We discuss the issues that are involved in formulating precursor versions of inference and then present some specific and highly visual proposals. These build on novel ways of experiencing sampling variation and have intuitive connections to the standard formal methods of making inferences in first university courses in statistics. Our proposal uses visual comparisons to enable the inferential step to be made without taking the eyes off relevant graphs of the data. This allows the time and conceptual distances between questions, data and conclusions to be minimized, so that the most critical linkages can be made. Our approach was devised for use in high schools but is also relevant to adult education and some introductory tertiary courses.
 
 
To be published in Series A.
 
Download: