Date: May 13-15, 16.00-17.30

Venue: Garching-Hochbrueck, Parkring 13, Room Number 2.02.01

Summary:

With the advent of widely accessible powerful computing in the late 1970s, computer-intensive methods have created a revolution in modern statistics. In particular, resampling methods such as the bootstrap have been developed to obtain powerful solutions to statistical inference. However, even when the model assumptions are correct, bootstrap prediction intervals for regression (and autoregression) have been plagued by undercoverage; this is true even in the simplest case of linear regression. Furthermore, in many instances the model assumptions can be violated in which case any model-based inference will be invalid.

In this series of seminars, the problem of statistical prediction will be revisited with a view that goes beyond the typical parametric/nonparametric dilemmas in order to reach a fully model-free environment for predictive inference, i.e., point predictors and predictive intervals. The `Model-Free (MF) Prediction Principle' of
Politis (2007) is based on the notion of transforming a given set-up into one that is easier to work with, namely i.i.d. or Gaussian. The two important applications are regression and autoregression whether an additive parametric/nonparametric model is applicable or not.

Requirements:
This seminar series is designed for students (interested Bachelor and Master students), PhD students, postdoctoral fellows and colleagues from basically all fields with background in regression methods in statistics; some familiarity with time series analysis would be helpful in connection with the 3rd seminar.

Material:
D.N. Politis, `Model-free prediction',
Bulletin of the International Statistical Institute,
Volume LXII , Lisbon, 2007, pp. 1391-1397.
Available from:
http://www.math.ucsd.edu/~politis/PAPER/MF2.pdf