[Statlist] talks on statistics

[Susanne Kaiser-Heinzmann]

                 ETH and University of Zurich 

                           Proff. 
         A.D. Barbour - P. Buehlmann - F. Hampel - L. Held

           H.R. Kuensch - M. Maathuis - S. van de Geer


***********************************************************    
           We are glad to announce the following talks
***********************************************************
         Friday, May 9, 2008, 15.15-17.00, LEO C6 

         Estimation of Optimal Dynamic Anticoagulation Regimes from
         Observational Data: A Regret-Based Approach
         Robin Henderson, Newcastle University, UK

Dynamic  regimes are employed when decisions have to be made reactively 
as data becomes available though time. As an example consider the 
problem of dosage selection for any of the two million European patients 
who are provided with long term anticoagulation.  Optimal dose does not 
just vary between patients, it varies within patients over time in 
response to short-term changes in lifetsyle or diet.  At a clinic visit 
at timepoint t the physician needs to review the state S(t) of the 
patient and then make a decision as to what what action A(t), is needed, 
such as what dose to prescribe.  We assume that the goal of the 
actions/treatments is to maximise some final quantity Y.

Dynamic programming methodology provides the traditional approach to 
determining decision rules to optimise Y.  This requires a model of the 
consequences which an action at each t will have on both the final value 
Y and all interim states {S(t+u)}.  This is a direct approach. However, 
there are many computational problems unless there are only low  numbers 
of possible states and actions.

An alternative approach, designed to be applicable to complex 
observational data, is to avoid modelling the direct consquences of an 
action but instead to take an assumed parametric form for the difference 
in expected final outcomes Y given two possible decisions (Moodie et al, 
Biometrics 2007).  Modelling and estimation under such an indirect 
approach is described in this talk, and an application on 
antocoagulation is presented.

***********************************************************
            
         Friday, May 16, 2008, 15.15-17.30, LEO C6, 
         Non-asymptotic variable identification via the Lasso 
         and the elastic net
         Florentina Bunea, Florida State University  

  The topic of $\ell_1$ regularized or Lasso type estimation has received
  considerable attention over the past decade. Recent theoretical advances
  have been mainly concerned with the risk of the estimators and
  corresponding sparsity oracle inequalities. In this talk we will
  investigate the quality of the $\ell_1$ penalized estimators from a
  different perspective, shifting the emphasis to non-asymptotic variable
  selection, which complements the consistent variable selection
  literature. Our main results are established for regression models, with
  emphasis on the square and logistic loss. The identification of the
  tagged SNPs associated with a disease, in genome wide association
  studies, provides the principal motivation for this analysis. The
  performance of the method depends crucially on the choice of the tuning
  sequence and we discuss non-asymptotic choices for which we can correctly
  detect sets of variables associated with the response at any
  pre-specified confidence level. These tuning sequences are different for
  the two loss functions, but in both cases larger than those required for
  best risk performance. The stability of the design matrix is another
  major issue in correct variable selection, especially when the total
  number of variables exceeds the sample size. A possible solution is
  provided by further regularization, for instance via an $\ell_1 + \ell_2$
  or elastic net type penalty. We discuss the merits and limitations of
  this method in the same context as above.
*********************************************************** 
 
_______________________________________________________
Christina Kuenzli            <kuenzli at stat.math.ethz.ch>
Seminar fuer Statistik      
Leonhardstr. 27,  LEO D11      phone: +41 (0)44 632 3438         
ETH-Zentrum,                   fax  : +41 (0)44 632 1228 
CH-8092 Zurich, Switzerland        http://stat.ethz.ch/~