Irène Gijbels

Edge-preserving image denoising and estimation of discontinuous surfaces

In this paper, we are interested in the problem of estimating a discontinuous surface from noisy data. A novel procedure for this problem is proposed based on local linear kernel smoothing, in which local neighborhoods are adapted to the local smoothness of the surface measured by the observed data. The procedure can therefore remove noise correctly in continuity regions of the surface and preserve discontinuities at the same time. Since an image can be regarded as a surface of the image intensity function and such a surface has discontinuities at the outlines of objects, this procedure can be applied directly to image denoising. Numerical studies show that it works well in applications, compared to some existing procedures

Semiparametric estimation of conditional copulas

The manner in which two random variables influence one another often depends on covariates. A way to model this dependence is via a conditional copula function. This paper contributes to the study of semiparametric estimation of conditional copulas by starting from a parametric copula function in which the parameter varies with a covariate, and leaving the marginals unspecified. Consequently, the unknown parts in the model are the parameter function and the unknown marginals. The authors use a local pseudo-likelihood with nonparametrically estimated marginals approximating the unknown parameter function locally by a polynomial. Under this general setting, they prove the consistency of the estimators of the parameter function as well as its derivatives; they also establish asymptotic normality. Furthermore, they derive an expression for the theoretical optimal bandwidth and discuss practical bandwidth selection. They illustrate the performance of the estimation procedure with data-driven bandwidth selection via a simulation study and a real-data case.

Variable Selection in Additive Models Using P-Splines

This article extends the nonnegative garrote method to a component selection method in a nonparametric additive model in which each univariate function is estimated with P-splines. We also establish the consistency of the procedure. An advantage of P-splines is that the fitted function is represented in a rather small basis of B-splines. A numerical study illustrates the finite-sample performance of the method and includes a comparison with other methods. The nonnegative garrote method with P-splines has the advantage of being computationally fast and performs, with an appropriate parameter selection procedure implemented, overall very well. Real data analysis leads to interesting findings. Supplementary materials for this article (technical proofs, additional numerical results, R code) are available online.

Variable Selection in Varying-Coefficient Models Using P-Splines

In this article, we consider nonparametric smoothing and variable selection in varying-coefficient models. Varying-coefficient models are commonly used for analyzing the time-dependent effects of covariates on responses measured repeatedly (such as longitudinal data). We present the P-spline estimator in this context and show its estimation consistency for a diverging number of knots (or B-spline basis functions). The combination of P-splines with nonnegative garrote (which is a variable selection method) leads to good estimation and variable selection. Moreover, we consider APSO (additive P-spline selection operator), which combines a P-spline penalty with a regularization penalty, and show its estimation and variable selection consistency. The methods are illustrated with a simulation study and real-data examples. The proofs of the theoretical results as well as one of the real-data examples are provided in the online supplementary materials.

Frequent problems in calculating integrals and optimizing objective functions: a case study in density deconvolution

Abstract Many statistical procedures involve calculation of integrals or optimization (minimization or maximization) of some objective function. In practical implementation of these, the user often has to face specific problems such as seemingly numerical instability of the integral calculation, choices of grid points, appearance of several local minima or maxima, etc. In this paper we provide insights into these problems (why and when are they happening?), and give some guidelines of how to deal with them. Such problems are not new, neither are the ways to deal with them, but it is worthwhile to devote serious considerations to them. For a transparant and clear discussion of these issues, we focus on a particular statistical problem: nonparametric estimation of a density from a sample that contains measurement errors. The discussions and guidelines remain valid though in other contexts. In the density deconvolution setting, a kernel density estimator has been studied in detail in the literature. The estimator is consistent and fully data-driven procedures have been proposed. When implemented in practice however, the estimator can turn out to be very inaccurate if no adequate numerical procedures are used. We review the steps leading to the calculation of the estimator and in selecting parameters of the method, and discuss the various problems encountered in doing so.

Data-driven boundary estimation in deconvolution problems

Estimation of the support of a density function is considered, when only a contaminated sample from the density is available. A kernel-based method has been proposed in the literature, where the authors study theoretical bias and variance of the estimator. Practical implementation issues of this method are considered here, which are a necessary supplement to the theoretical results to get to a data-driven method that is widely applicable. Two such practical data-driven procedures are proposed. Simulation results show that they perform well for a wide variety of densities (including quite difficult cases). The methods can also be applied for error-free data and as such also present data-driven procedures for estimation of boundaries in the case of non-contaminated data. Moreover they can be applied for estimating discontinuities of a density, as is shown. The proposed data-driven boundary estimation procedures are illustrated in frontier estimation.

Robust Estimation of Mean and Dispersion Functions in Extended Generalized Additive Models

Generalized linear models are a widely used method to obtain parametric estimates for the mean function. They have been further extended to allow the relationship between the mean function and the covariates to be more flexible via generalized additive models. However, the fixed variance structure can in many cases be too restrictive. The extended quasilikelihood (EQL) framework allows for estimation of both the mean and the dispersion/variance as functions of covariates. As for other maximum likelihood methods though, EQL estimates are not resistant to outliers: we need methods to obtain robust estimates for both the mean and the dispersion function. In this article, we obtain functional estimates for the mean and the dispersion that are both robust and smooth. The performance of the proposed method is illustrated via a simulation study and some real data examples.

Robust nonnegative garrote variable selection in linear regression

Robust selection of variables in a linear regression model is investigated. Many variable selection methods are available, but very few methods are designed to avoid sensitivity to vertical outliers as well as to leverage points. The nonnegative garrote method is a powerful variable selection method, developed originally for linear regression but recently successfully extended to more complex regression models. The method has good performances and its theoretical properties have been established. The aim is to robustify the nonnegative garrote method for linear regression as to make it robust to vertical outliers and leverage points. Several approaches are discussed, and recommendations towards a final good performing robust nonnegative garrote method are given. The proposed method is evaluated via a simulation study that also includes a comparison with existing methods. The method performs very well, and often outperforms existing methods. A real data application illustrates the use of the method in practice.

P-splines regression smoothing and difference type of penalty

P-splines regression provides a flexible smoothing tool. In this paper we consider difference type penalties in a context of nonparametric generalized linear models, and investigate the impact of the order of the differencing operator. Minimizing Akaike’s information criterion we search for a possible best data-driven value of the differencing order. Theoretical derivations are established for the normal model and provide insights into a possible ‘optimal’ choice of the differencing order and its interrelation with other parameters. Applications of the selection procedure to non-normal models, such as Poisson models, are given. Simulation studies investigate the performance of the selection procedure and we illustrate its use on real data examples.

Weak convergence of discretely observed functional data with applications

A general result on weak convergence of the empirical measure of discretely observed functional data is shown. It is applied to the problem of estimation of functional mean value, and the problem of consistency of various types of depth for functional data. Counterexamples illustrating the fact that the assumptions as stated cannot be dropped easily are given.