Tutorial, Bayes net for forensic DNA analysis, Satellite, Dice, DNA Helix, Distance Matrix of pairs of amino acids, Bristol Balloon festival
SuSTaIn About News Events Closing workshop... EC: High-dim Statistics... EC: Hidden Complexities Intractable Likelihoods UK Causal Inference Meeting EC: Astrostatistics Image Processing Extremes Structure and Uncertainty Functional data analysis Confronting Intractability Time for Causality Laurie Davies lectures Julian Besag memorial Research highlights Jobs Management Statistics Group Statistics Home Research Members Seminars Mathematics Home External Links APTS Complexity science Royal Statistical Society International Society for Bayesian Analysis


Approximating data

Professor Laurie Davies (University of Duisburg-Essen)

Professor Davies will present a series of 3 lectures, between 12 and 17 October 2011.

Location: School of Mathematics, University of Bristol: see the university precinct map (select 'Mathematics' from the key).

Note that the times and precise locations vary from day to day.

  • Wednesday 12 October, 2pm, room SM3: Approximation and Approximation Regions
  • Friday 14 October, 3pm, room SM3: Regularization
  • Monday 17 October, 11am, room SM4: On some statistical concepts

Slides for lecture 1 | lecture 2 | lecture 3


The three talks describe and develop a way of approaching statistics whereby models are consistently considered as approximations to the data: a model P is an adequate approximation to a data set x_n if typical data sets generated under P look like the data x_n. Informally this means that a statistician confronted with simulated samples and the real sample cannot reliably identify the real sample. The basic idea is simple but it has consequences for many common concepts in statistics such as likelihood and efficiency. Most of the examples of the use of the idea come from the area of non-parametric regression.