Contact

Email: Sylvie.Huet AT jouy.inra.fr Mailing address: INRA UNITÉ MIA Domaine de Vilvert 78352 Jouy-en-Josas Cedex - FRANCE Tél: (+33) (0)1 34 65 22 24 Fax: (+33) (0)1 34 65 22 17 Secretary: (+33) (0)1 34 65 22 31

Research Topics

I am working on statistical models usefull for analysing data observed by biologists working at INRA.

• More particularly I studied nonlinear regression models with heteroscedastic variances and non parametric regression models proposing methods for inference based on bootstrap. These models are widely used for estimating dose response relationships.
• I also proposed goodness of fit tests for testing the validity of a regression model.
• More recently I worked on model selection procedures for detecting non-zero components in a gaussian vector, for estimating change points in a signal or for selecting variables among a large set of variables.
• At present, I consider models where the number of parameters to estimate is large face to the number of obervations, focusing on inference of graphical gaussian models, of relations between variables.

Applications

• I contribute to "estimating landscape gene flow" through my participation to a project studying the origin of feral oilseed rape populations. I used mixed logistic models for this study and data mining procedures.
• I am involved in the analysis of post-genomic data : comparison of protein abundance on the basis of two-dimensional electrophoresis images, estimating interactions of cellular components on the basis of "omic" data, estimating relations between phenotypes and "omic" data.

Publications

Book

Technical reports

Graph selection with GGMselect : with C. Giraud and N. Verzelen

Estimator selection in the Gaussian setting : with Y. Baraud and C. Giraud.

Software

NLS2 : A set of Splus (or R) functions and programs to estimate the parameters of a non-linear regression model over a given set of observations.

GGMselect : A R package dedicated to graph estimation in Gaussian Graphical Models. The main functions return the adjacency matrix of an undirected graph estimated from a data matrix.

LINSelect : R functions for choosing an estimator in Gaussian regression model according to the method proposed by Baraud et al. 2010. The functions are available in a zip file.

TuningLasso : R functions for choosing the tuning parameter in the Lasso procedure : Vfold CV, LinSelect, square-root Lasso (also known as scaled-Lasso). Moreover Vfold CV and LinSelect are available for tuning the parameter of the Gauss-Lasso procedure. Giraud et al. 2011. The functions are available in a zip file.

ConvDiscr : R functions for the non prametric estimation of a convex discrete distribution. Durot et al. 2012. The functions are available in a zip file.