Search

This article studies M-type estimators for fitting robust additive models in the presence of anomalous data. The components in the additive model are allowed to have different degrees of smoothness. We introduce a new class of wavelet-based robust M-type estimators for performing simultaneous additive component estimation and variable selection in such inhomogeneous additive models. Each additive component is approximated by a truncated series expansion of wavelet bases, making it feasible to apply the method to nonequispaced data and sample sizes that are not necessarily a power of 2. Sparsity of the additive components together with sparsity of the wavelet coefficients within each component (group), results into a bi-level group variable selection problem. In this framework, we discuss robust estimation and variable selection. A two-stage computational algorithm, consisting of a fast accelerated proximal gradient algorithm of coordinate descend type, and thresholding, is proposed. When using nonconvex redescending loss functions, and appropriate nonconvex penalty functions at the group level, we establish optimal convergence rates of the estimates. We prove variable selection consistency under a weak compatibility condition for sparse additive models. The theoretical results are complemented with some simulations and real data analysis, as well as a comparison to other existing methods.

Facilitating the identification of extreme inactivity (EI) has the potential to improve morbidity and mortality in COPD patients. Apart from patients with obvious EI, the identification of a such behavior during a real-life consultation is unreliable. We therefore describe a machine learning algorithm to screen for EI, as actimetry measurements are difficult to implement. Complete datasets for 1409 COPD patients were obtained from COLIBRI-COPD, a database of clinicopathological data submitted by French pulmonologists. Patient- and pulmonologist-reported estimates of PA quantity (daily walking time) and intensity (domestic, recreational, or fitness-directed) were first used to assign patients to one of four PA groups (extremely inactive [EI], overtly active [OA], intermediate [INT], inconclusive [INC]). The algorithm was developed by (i) using data from 80% of patients in the EI and OA groups to identify ‘phenotype signatures’ of non-PA-related clinical variables most closely associated with EI or OA; (ii) testing its predictive validity using data from the remaining 20% of EI and OA patients; and (iii) applying the algorithm to identify EI patients in the INT and INC groups. The algorithm’s overall error for predicting EI status among EI and OA patients was 13.7%, with an area under the receiver operating characteristic curve of 0.84 (95% confidence intervals: 0.75–0.92). Of the 577 patients in the INT/INC groups, 306 (53%) were reclassified as EI by the algorithm. Patient- and physician- reported estimation may underestimate EI in a large proportion of COPD patients. This algorithm may assist physicians in identifying patients in urgent need of interventions to promote PA.

Jonathan Lévesque,1 Anestis Antoniadis,2 Pei Zhi Li,3 Frédéric Herengt,4 Christophe Brosson,5 Jean-Marie Grosbois,6 Alain Bernady,7 Anthony Bender,8,9 Murielle Favre,10 Antoine Guerder,11 Pascale Surpas,12 Thomas Similowski,11,13 Bernard Aguilaniu141Department of Medicine, Pneumology Service, Hôpital Maisonneuve-Rosemont, Faculty of Medicine, Université de Montréal, Montreal, Quebec, Canada; 2Laboratoire Jean Kuntzmann UMR5224, Statistics Department, Université Grenoble-Alpes, Grenoble, France; 3Montreal Chest Institute, McGill University Health Center Research Institute, McGill University, Montreal, Quebec, Canada; 4Centre de Réadaptation Cardio-Respiratoire-Dieulefit Santé, Dieulefit, France; 5Private Physiotherapist, Cours Liberation, Grenoble, France; 6Pneumology Service, Centre Hospitalier de Béthune, Béthune, France; 7Toki-Eder Centre Médical Cardio-Respiratoire, Cambo-les-Bains, France; 8Private Physiotherapist, Boulevard de Metz Mont-Saint-Martin, France; 9LUNEX International University of Health, Exercise and Sports, Differdange, Luxembourg; 10Centre de Pneumologie Henri Bazire, Saint-Julien-de-Ratz, France; 11AP-HP, Groupe Hospitalier Pitié-Salpêtrière Charles Foix, Service de Pneumologie, Médecine Intensive et Réanimation, Département R3S, Paris, France; 12Centre Médical de Bayère, Charnay, France; 13Sorbonne Université, INSERM, UMRS1158 Neurophysiologie Respiratoire Expérimentale et Clinique, Paris, France; 14Faculty of Medicine, Université Grenoble-Alpes, Grenoble, FranceBackground: The 3-minute chair rise test (3-minute CRT) and the Disability Related to COPD Tool (DIRECT) are two reproducible and valid short tests that can assess the benefit of pulmonary rehabilitation (PR) in terms of functional capacity and dyspnea in everyday activities.Methods: We determined the minimal clinically important difference (MCID) of the DIRECT questionnaire and 3-minute CRT using distribution methods and anchor encroaches with a panel of eight standard tests in a cohort of 116 COPD patients who completed a PR program in real-life settings.Results: The estimated MCID for the 3-minute CRT and DIRECT scores was five repetitions and two units, respectively, using separate and combined independent anchors. The all-patient (body mass index-obstruction-dyspnea-exercise [BODE] scores 0–7), BODE 0–2 (n=42), and BODE 3–4 (n=50) groups showed improvements greater than the MCID in most tests and questionnaires used. In contrast, the BODE 5–7 group (n=24) showed improvements greater than MCID in only the 3-minute CRT, 6-minute walk test, endurance exercise test, and DIRECT questionnaire.Discussion and conclusion: This study demonstrates that the short and simple DIRECT questionnaire and 3-minute CRT are responsive to capture the beneficial effects of a PR program in COPD patients, including those with severe disease.Trial registration number: NCT03286660. Keywords: MCID, chair tests, tools, COPD, pulmonary rehabilitation, outcomes assessment

We introduce a new class of robust M-estimators for performing simultaneous parameter estimation and variable selection in high-dimensional regression models. We first explain the motivations for the key ingredient of our procedures which are inspired by regularization methods used in wavelet thresholding in noisy signal processing. The derived penalized estimation procedures are shown to enjoy theoretically the oracle property both in the classical finite dimensional case as well as the high-dimensional case when the number of variables p is not fixed but can grow with the sample size n, and to achieve optimal asymptotic rates of convergence. A fast accelerated proximal gradient algorithm, of coordinate descent type, is proposed and implemented for computing the estimates and appears to be surprisingly efficient in solving the corresponding regularization problems including the case for ultra high-dimensional data where $$p \gg n$$ . Finally, a very extensive simulation study and some real data analysis, compare several recent existing M-estimation procedures with the ones proposed in the paper, and demonstrate their utility and their advantages.

Short-Term Load Forecasting (STLF) is a fundamental instrument in the efficient operational management and planning of electric utilities. Emerging smart grid technologies pose new challenges and opportunities. Although load forecasting at the aggregate level has been extensively studied, electrical load forecasting at fine-grained geographical scales of households is more challenging. Among existing approaches, semi-parametric generalized additive models (GAM) have been increasingly popular due to their accuracy, flexibility, and interpretability. Their applicability is justified when forecasting is addressed at higher levels of aggregation, since the aggregated load pattern contains relatively smooth additive components. High resolution data are highly volatile, forecasting the average load using GAM models with smooth components does not provide meaningful information about the future demand. Instead, we need to incorporate irregular and volatile effects to enhance the forecast accuracy. We focus on the analysis of such hybrid additive models applied on smart meters data and show that it leads to improvement of the forecasting performances of classical additive models at low aggregation levels. (C) 2020 International Institute of Forecasters. Published by Elsevier B.V. All rights reserved.

Background: In idiopathic pulmonary fibrosis (IPF), some physiological parameters measured during a 6-min walk test (6-MWT) impart reliable prognostic information. Sit-to-stand tests (STSTs) are field exercise tests that are easier to implement than the 6-MWT in daily practice. Objectives: The aims of the study were to test the reproducibility and compare 2 STSTs (the 1-min STST [1-STST] and the semi-paced 3-min chair rise test [3-CRT]) in IPF, and to determine if selected physiological parameters (speed of displacement and changes in pulse oxygen saturation [SpO2]) are interchangeable between the STSTs and the 6-MWT. Methods: Thirty-three patients with stable IPF were studied in 3 French expert centers. To test reproducibility, intra-class correlations (ICCs) of parameters measured during tests performed 7–14 days apart were calculated. To test interchangeability, the agreement and correlation of physiological responses measured during STSTs and during 6-MWT were studied. Results: Vertical displacements and changes in SpO2 during both STSTs were reproducible, with ICCs ranging from 0.78 [0.63–0.87] to 0.95 [0.92–0.97]. Vertical displacements during 1-STST and 3-CRT were correlated with 6-MWT distance (correlation coefficients (r) of 0.72 and 0.77, respectively; p < 0.001). Similarly, correlations were found between changes in SpO2 measured during the 2 STSTs and the 6-MWT, with coefficients ranging from 0.73 to 0.91 (p < 0.001). Distance walked and SpO2 during 6-MWT were well estimated from vertical displacement and SpO2 during the 2 STSTs, respectively. Conclusion: The correlations found between the 2 STSTs and the 6-MWT suggest that STSTs may be of interest to assess displacement and exercise-induced changes in SpO2 in IPF patients.

Based on the hypothesis that many continuous and categorical variables collected during the consultation could be affected (as cause or consequence) by excessively sedentary and physically inactive behaviour (ESPI), we developed a predictive method based on all the information collected during the COPD digital consultation (n=5040). The variable to be predicted, i.e. the ESPI, was defined by crossing 2 estimates among several categories of activities proposed to the doctor and the patient. At the other end of the spectrum, we defined also an ACTIVE status. Methods: we first verified that the ESPI variable and its variability were well correlated with a set of continuous and categorical variables measured by performing a factor analysis on mixed data. A scree plot was then used to determine the statistically significant factors of the components to be included in more elaborate individual predictive models. To ensure the robustness of the prediction, we compared two predictive models (a multiple logistic regression and a random forest), and to generalize their predictive power through estimates of the uncertainty of their predictions, these models were improved by sequentially adding several random effects such as the identity of the doctor, the hospital centre, etc. The models were then used to estimate the uncertainty of the predictions. Results: The set of methods used allows to predict the ESPI and ACTIVE status with an error rate of 33% and 7% respectively. Conclusion: This proof-of-concept study demonstrates the ability to detect ESPI status using a machine learning powered by common variables from a digital consultation including a wide range of COPD patients.

A Round Robin exercise was implemented by ESA to compare different classification methods in detecting clouds from images taken by the PROBA-V sensor. A high-quality dataset of 1350 reflectances and Clear/Cloudy corresponding labels had been prepared by ESA in the framework of the exercise. Motivated by both the experience acquired by one of the authors in this exercise and the availability of such a reliable annotated dataset, we present a full assessment of the methodology proposed therein. Our objective is also to investigate specific issues related to cloud detection when remotely sensed images comprise only a few spectral bands in the visible and near-infrared. For this purpose, we consider a bunch of well-known classification methods. First, we demonstrate the feasibility of using a training dataset semi-automatically obtained from other accurate algorithms. In addition, we investigate the effect of ancillary information, e.g., surface type or climate, on accuracy. Then we compare the different classification methods using the same training dataset under different configurations. We also perform a consensus analysis aimed at estimating the degree of mutual agreement among classification methods in detecting Clear or Cloudy sky conditions.

Introduction L’evaluation de la rehabiltation (RHB) s’appuie sur le changement significatif de scores (Dyspnee-QV) et de grandeurs physiologiques d’exercice. Nous avons etabli le MCID du TLC3’ et du questionnaire DIRECT au cours d’une etude en vie reelle [1] . Au cours de cette etude, les tests physiologiques ont ete repetes apres 6 seances d’entrainement afin de juger si l’ampleur du benefice final etait en partie due a l’amelioration de l’aisance loco-motrice. Methodes 116 patients GOLD II, III, IV (VEMS moyen: 1,3 ± 0.5 L; 48 ± 16 % pred.) ont ete evalues en debut et fin de programme mais aussi precocement (a l’exception des questionnaires) apres 6 a 8 seances soit en moyenne au 14e jour (11,8–16,4). Duree moyenne du programme: 42 jours. Les tests d’exercice etaient: TM6, Temps endurance Cycloergometre, TLC-3’, TLC-1’, force musculaire quadriceps (n = 38). Les questionnaires (debut et fin de programme etaient: mMRC, CAT, SGRQ, DIRECT, HAD. Resultats Une amelioration significative etait atteinte pour tous les tests d’exercice TM6, Endurance Cycle, TLC3, TLC1, FMQD, FLQG (respectivement de 48,9 m, 4,1 min, 9,1 levers, 5,7 levers, 13,5, 15,4 U.); et pour tous les scores (excepte le mMRC) soit DIRECT, CAT, SGRQ total respectivement −1,9, −2,7, − 4,7 Unites. Des la 6e seance, environ 50 % du benefice final etait constate pour TM6, Endurance Cycle, TLC3, TLC1, FMQ atteignant le MCID pour TLC3-TLC1-Endurance cycle. Conclusion La RHB est jugee efficace quel que soit le stade GOLD selon le test realise attestant la necessite d’une evaluation multi dimensionnelle avec des tests courts, facilement repetables et avec un MCID valide [2] . Les tests simples s’affranchissant des contraintes liees a l’environnement de travail sont aussi performants que les tests complexes (TM6-Endurance cyclo) pour juger du benefice de la RHB. L’evaluation du benefice physiologique reel d’un programme d’entrainement devrait prendre pour comparaison les mesures fonctionnelles initiales realisees apres 6-8 seances d’entrainement.

Introduction L’interrogatoire aide par des questionnaires specifique (QV-Dyspnee) permet de distinguer assez facilement les patients BPCO excessivement inactifs (EI) de ceux qui maintiennent une activite physique suffisante. Entre ces deux situations le doute subsiste ou n’est pas evoque, alors que les etudes s’accordent pour detecter, grâce aux mesures d’accelerometrie, plus de 40 % de patients inactifs dans cette sous population independamment de la severite de l’obstruction bronchique. Partant de l’hypothese que l’EI pouvait impacter, comme cause ou consequence, de multiples variables recueillies en consultation de routine, nous avons verifie si le comportement EI pouvait etre predit individuellement a partir de ces variables excluant de fait toutes les mesures d’exercice realisees au-dela du temps de la consultation. Methodes A partir de 5035 consultations Colibri-BPCO, nous avons selectionne 1409 dossiers ayant des informations completes sur 22 variables d’interet. Nous avons determine sur une base clinique un groupe de 172 patients denome SBPI associant un comportement sedentaire (SB) et physiquement inactif (PI) et un autre de 660 patients qui ne l’etait pas (non-SBPI). Sur les 832 patients SBPI et non SBPI, les donnees de 80 % (n = 666) ont ete utilisees pour mettre au point un algorithme d’apprentissage automatique pour predire le statut SBPI, et les donnees des 20 % (n = 166) restants ont ete utilisees pour tester sa fiabilite, (c’est dire l’erreur de prediction). L’algorithme a ensuite ete applique pour predire le statut SBPI des 577 patients « intermediaires » (1409–[172 + 660]). Resultats L’erreur globale de l’algorithme de prediction du statut SBPI etait de 13,7 %, avec une aire sous la courbe ROC de 0,84 (intervalles de confiance de 95 %: 0.75–0.92). Sur les 577 patients de la categorie intermediaire, 306 (53 %) ont ete classes comme SBPI par l’algorithme, ce qui suggere que les patients et/ou les medecins ont subjectivement a surestime les veritables niveaux d’AP des patients. Conclusion Cette etude demontre la fiabilite et l’utilite potentielle d’un algorithme predictif pour identifier a partir d’une consultation de routine, sans aucune mesure d’activite ou de capacite physique, si un patient presente une inactivite morbide necessitant une intervention prioritaire.

French electricity load forecasting has encountered major changes during the past decade. These changes are, among other things, due to the opening of the electricity market and the economic crisis, which require the development of new automatic time adaptive prediction methods. The advent of innovating technologies also needs the development of some automatic methods because thousands or tens of thousands of time series have to be studied. In this paper we adopt for prediction a semi-parametric approach based on additive models. We present an automatic procedure for explanatory variable selection in an additive model and show how to correct middle term forecasting errors for short term forecasting. First, we consider an application to the EDF customer load demand which is typical of a load demand at an aggregate level. The goal of the application is to select variables from a large explanatory variables dictionary. The second application presented is an application on load demand of GEFCom 2012 competition, which we consider as a local application, where a major difficulty is to select some meteorological stations.

Starting from the information contained in the shape of the load curves, we propose a flexible nonparametric function-valued forecast model called KWF (Kernel + Wavelet + Functional) that is well suited to the handling of nonstationary series. The predictor can be seen as a weighted average of the futures of past situations, where the weights increase with the similarity between the past situations and the actual one. In addition, this strategy also provides simultaneous predictions at multiple horizons. These weights induce a probability distribution that can be used to produce bootstrap pseudo predictions. Prediction intervals are then constructed after obtaining the corresponding bootstrap pseudo prediction residuals. We develop two propositions following the KWF strategy directly, and compare it to two alternative methods that arise from proposals by econometricians. The latter involve the construction of simultaneous prediction intervals using multiple comparison corrections through the control of the family-wise error (FWE) or the false discovery rate. Alternatively, such prediction intervals can be constructed by bootstrapping joint probability regions. In this work, we propose to obtain prediction intervals for the KWF model that are valid simultaneously for the H prediction horizons that correspond to the relevant path forecasts, making a connection between functional time series and the econometricians’ framework.

We congratulate the authors for a very interesting and timely contribution to an important problem: using nonparametric functional data analysis methods for studying spatio-temporal observations with the aim to identify subregions in a spatial domain sharing similar patterns along time. The authors develop an efficient methodology to perform dimensional reduction of spatially dependent functional data based on treelet decompositions and an appropriate bagging Voronoi strategy that allows them to take into account the spatial dependency in the analysed data. Treelets were first introduced in Lee and Nadler (2007) and Lee et al. (2008) as a multi-scale basis that extends wavelets to unordered data. The method is fully adaptive. It returns orthonormal basis functions supported on nested clusters in a hierarchical tree. Unlike other hierarchical methods, the basis and the tree structure are computed simultaneously, and both reflect the internal structure of the data. Combination of treelets with a bagging Voronoi strategy for identifying subregions with similar patterns along time introduces numerical challenges that are efficiently tackled by the authors.

In this paper we study sparse high dimensional additive partial linear models with nonparametric additive components of heterogeneous smoothness. We review several existing algo- rithms that have been developed for this problem in the recent literature, highlighting the connec- tions between them, and present some computationally efficient algorithms for fitting such models. To achieve optimal rates in large sample situations we use hybrid P-splines and block wavelet penal- isation techniques combined with adaptive (group) LASSO-like procedures for selecting the additive components in the nonparametric part of the models. Hence, the component selection and estimation in the nonparametric part may be viewed as a functional version of estimation and grouped variable selection. This allows to take advantage of several oracle results which yield asymptotic optimality of estimators in high-dimensional but sparse additive models. Numerical implementations of our procedures for proximal like algorithms are discussed. Large sample properties of the estimates and of the model selection are presented and the results are illustrated with simulated examples and a real data analysis.

Given two independent samples of size n and m drawn from univariate distributions with unknown densities f and g, respectively, we are interested in identifying subintervals where the two empirical densities deviate significantly from each other. The solution is built by turning the nonparametric density comparison problem into a comparison of two regression curves. Each regression curve is created by binning the original observations into many small size bins, followed by a suitable form of root transformation to the binned data counts. Turned as a regression comparison problem, several nonparametric regression procedures for detection of sparse signals can be applied. Both multiple testing and model selection methods are explored. Furthermore, an approach for estimating larger connected regions where the two empirical densities are significantly different is also derived, based on a scale-space representation. The proposed methods are applied on simulated examples as well as real-life data from biology.

We consider estimation and model selection in sparse high-dimensional linear additive models when multiple covariates need to be modeled nonparametrically, and propose some multi-step estimators based on B-splines approximations of the additive components. In such models, the overall number of regressors d can be large, possibly much larger than the sample size n. However, we assume that there is a smaller than n number of regressors that capture most of the impact of all covariates on the response variable. Our estimation and model selection results are valid without assuming the conventional “separation condition”—namely, without assuming that the norm of each of the true nonzero components is bounded away from zero. Instead, we relax this assumption by allowing the norms of nonzero components to converge to zero at a certain rate. The approaches investigated in this paper consist of two steps. The first step implements the variable selection, typically by the Group Lasso, and the second step applies a penalized P-splines estimation to the selected additive components. Regarding the model selection task we discuss, the application of several criteria such as Akaike information criterion (AIC), Bayesian information criterion (BIC), and generalized cross validation (GCV) and study the consistency of BIC, i.e. its ability to select the true model with probability converging to 1. We then study post-model estimation consistency of the selected components. We end the paper by applying the proposed procedure on some real data related to electricity load consumption forecasting: the EDF (Electricite de France) portfolio.

© 2016 Institute of Mathematical Statistics. All rights reserved. When describing adequately complex data structures one is often confronted with the fact that mean as well as variance (or more generally dispersion) is highly influenced by some covariates. Drawbacks of the available methods is that they are often based on approximations and hence a theoretical study should deal with also studying these approximations. This however is often ignored, making the statistical inference incomplete. In the proposed framework of double generalized modelling based on proper dispersion models we avoid this drawback and as such are in a good position to use recent results on Bregman divergence for establishing theoretical results for the proposed estimators in fairly general settings. We also study variable selection when there is a large number of covariates, with this number possibly tending to infinity with the sample size. The proposed estimation and selection procedure is investigated via a simulation study, that includes also a comparative study with competitors. The use of the methods is illustrated via some real data applications. ispartof: Electronic Journal of Statistics vol:10 issue:1 pages:1630-1676 status: published

International audience; 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.

In this paper, we propose a new approach for modeling and fitting high-dimensional response regression models in the setting of complex spatio-temporal dynamics. This study is motivated by investigating one of the major atmospheric phenomena which drives the rainfall regime in Western Africa: West African Monsoon. We are particularly interested in studying the influence of sea surface temperatures in the Gulf of Guinea on precipitation in Saharan and sub-Saharan.

The Dantzig selector (DS) is a recent approach of estimation in high-dimensional linear regression models with a large number of explanatory variables and a relatively small number of observations. As in the least absolute shrinkage and selection operator (LASSO), this approach sets certain regression coefficients exactly to zero, thus performing variable selection. However, such a framework, contrary to the LASSO, has never been used in regression models for survival data with censoring. A key motivation of this article is to study the estimation problem for Cox's proportional hazards (PH) function regression models using a framework that extends the theory, the computational advantages and the optimal asymptotic rate properties of the DS to the class of Cox's PH under appropriate sparsity scenarios. We perform a detailed simulation study to compare our approach with other methods and illustrate it on a well-known microarray gene expression data set for predicting survival from gene expressions.