Forecasting with missing data: Application to a real case

This paper presents a comparative analysis of linear and mixed modelsfor short term forecasting of a real data series with a high percentage of missing data. Data are the series of significant wave heights registered at regular periods of three hours by a buoy placed in the Bay of Biscay.The series is interpolated with a linear predictor which minimizes theforecast mean square error. The linear models are seasonal ARIMA models and themixed models have a linear component and a non linear seasonal component.The non linear component is estimated by a non parametric regression of dataversus time. Short term forecasts, no more than two days ahead, are of interestbecause they can be used by the port authorities to notice the fleet.Several models are fitted and compared by their forecasting behavior.

Temperature, air pollution, and mortality from myocardial infarction in São Paulo, Brazil

An increase in daily mortality from myocardial infarction has been observed in association with meteorological factors and air pollution in several cities in the world, mainly in the northern hemisphere. The objective of the present study was to analyze the independent effects of environmental variables on daily counts of death from myocardial infarction in a subtropical region in South America. We used the robust Poisson regression to investigate associations between weather (temperature, humidity and barometric pressure), air pollution (sulfur dioxide, carbon monoxide, and inhalable particulate), and the daily death counts attributed to myocardial infarction in the city of São Paulo in Brazil, where 12,007 fatal events were observed from 1996 to 1998. The model was adjusted in a linear fashion for relative humidity and day-of-week, while nonparametric smoothing factors were used for seasonal trend and temperature. We found a significant association of daily temperature with deaths due to myocardial infarction (P < 0.001), with the lowest mortality being observed at temperatures between 21.6 and 22.6ºC. Relative humidity appeared to exert a protective effect. Sulfur dioxide concentrations correlated linearly with myocardial infarction deaths, increasing the number of fatal events by 3.4% (relative risk of 1.03; 95% confidence interval = 1.02-1.05) for each 10 µg/m³ increase. In conclusion, this study provides evidence of important associations between daily temperature and air pollution and mortality from myocardial infarction in a subtropical region, even after a comprehensive control for confounding factors.

Changements dans la répartition des décès selon l'âge : une approche non paramétrique pour l'étude de la mortalité adulte

Au cours du siècle dernier, nous avons pu observer une diminution remarquable de la mortalité dans toutes les régions du monde, en particulier dans les pays développés. Cette chute a été caractérisée par des modifications importantes quant à la répartition des décès selon l'âge, ces derniers ne se produisant plus principalement durant les premiers âges de la vie mais plutôt au-delà de l'âge de 65 ans. Notre étude s'intéresse spécifiquement au suivi fin et détaillé des changements survenus dans la distribution des âges au décès chez les personnes âgées. Pour ce faire, nous proposons une nouvelle méthode de lissage non paramétrique souple qui repose sur l'utilisation des P-splines et qui mène à une expression précise de la mortalité, telle que décrite par les données observées. Les résultats de nos analyses sont présentés sous forme d'articles scientifiques, qui s'appuient sur les données de la Human Mortality Database, la Base de données sur la longévité canadienne et le Registre de la population du Québec ancien reconnues pour leur fiabilité. Les conclusions du premier article suggèrent que certains pays à faible mortalité auraient récemment franchi l'ère de la compression de la mortalité aux grands âges, ère durant laquelle les décès au sein des personnes âgées tendent à se concentrer dans un intervalle d'âge progressivement plus court. En effet, depuis le début des années 1990 au Japon, l'âge modal au décès continue d'augmenter alors que le niveau d'hétérogénéité des durées de vie au-delà de cet âge demeure inchangé. Nous assistons ainsi à un déplacement de l'ensemble des durées de vie adultes vers des âges plus élevés, sans réduction parallèle de la dispersion de la mortalité aux grands âges. En France et au Canada, les femmes affichent aussi de tels développements depuis le début des années 2000, mais le scénario de compression de la mortalité aux grands âges est toujours en cours chez les hommes. Aux États-Unis, les résultats de la dernière décennie s'avèrent inquiétants car pour plusieurs années consécutives, l'âge modal au décès, soit la durée de vie la plus commune des adultes, a diminué de manière importante chez les deux sexes. Le second article s'inscrit dans une perspective géographique plus fine et révèle que les disparités provinciales en matière de mortalité adulte au Canada entre 1930 et 2007, bien décrites à l'aide de surfaces de mortalité lissées, sont importantes et méritent d'être suivies de près. Plus spécifiquement, sur la base des trajectoires temporelles de l'âge modal au décès et de l'écart type des âges au décès situés au-delà du mode, les différentiels de mortalité aux grands âges entre provinces ont à peine diminué durant cette période, et cela, malgré la baisse notable de la mortalité dans toutes les provinces depuis le début du XXe siècle. Également, nous constatons que ce sont précisément les femmes issues de provinces de l'Ouest et du centre du pays qui semblent avoir franchi l'ère de la compression de la mortalité aux grands âges au Canada. Dans le cadre du troisième et dernier article de cette thèse, nous étudions la longévité des adultes au XVIIIe siècle et apportons un nouvel éclairage sur la durée de vie la plus commune des adultes à cette époque. À la lumière de nos résultats, l'âge le plus commun au décès parmi les adultes canadiens-français a augmenté entre 1740-1754 et 1785-1799 au Québec ancien. En effet, l'âge modal au décès est passé d'environ 73 ans à près de 76 ans chez les femmes et d'environ 70 ans à 74 ans chez les hommes. Les conditions de vie particulières de la population canadienne-française à cette époque pourraient expliquer cet accroissement.

Les inégalités sociales dans la durée de vie la plus commune : la répartition des décès selon l'âge et le quintile de défavorisation au Québec en 2000-2002 et 2005-2007

Nous avons choisi de focaliser nos analyses sur les inégalités sociales de mortalité spécifiquement aux grands âges. Pour ce faire, l'utilisation de l'âge modal au décès combiné à la dispersion des décès au-delà de cet âge s'avère particulièrement adapté pour capter ces disparités puisque ces mesures ne sont pas tributaires de la mortalité prématurée. Ainsi, à partir de la distribution des âges au décès selon le niveau de défavorisation, au Québec au cours des périodes 2000-2002 et 2005-2007, nous avons déterminé l'âge le plus commun au décès et la dispersion des durées de vie au-delà de celui-ci. L'estimation de la distribution des décès selon l'âge et le niveau de défavorisation repose sur une approche non paramétrique de lissage par P-splines développée par Nadine Ouellette dans le cadre de sa thèse de doctorat. Nos résultats montrent que l'âge modal au décès ne permet pas de détecter des disparités dans la mortalité des femmes selon le niveau de défavorisation au Québec en 2000-2002 et en 2005-2007. Néanmoins, on assiste à un report de la mortalité vers des âges plus avancés alors que la compression de la mortalité semble s'être stabilisée. Pour les hommes, les inégalités sociales de mortalité sont particulièrement importantes entre le sous-groupe le plus favorisé et celui l'étant le moins. On constate un déplacement de la durée de vie la plus commune des hommes vers des âges plus élevés et ce, peu importe le niveau de défavorisation. Cependant, contrairement à leurs homologues féminins, le phénomène de compression de la mortalité semble toujours s'opérer.

Estimation of dynamic latent variable models using simulated nonparametric moments

Given a model that can be simulated, conditional moments at a trial parameter value can be calculated with high accuracy by applying kernel smoothing methods to a long simulation. With such conditional moments in hand, standard method of moments techniques can be used to estimate the parameter. Since conditional moments are calculated using kernel smoothing rather than simple averaging, it is not necessary that the model be simulable subject to the conditioning information that is used to define the moment conditions. For this reason, the proposed estimator is applicable to general dynamic latent variable models. Monte Carlo results show that the estimator performs well in comparison to other estimators that have been proposed for estimation of general DLV models.

Estimation of dynamic latent variable models using simulated nonparametric moments

Abstract. Given a model that can be simulated, conditional moments at a trial parameter value can be calculated with high accuracy by applying kernel smoothing methods to a long simulation. With such conditional moments in hand, standard method of moments techniques can be used to estimate the parameter. Because conditional moments are calculated using kernel smoothing rather than simple averaging, it is not necessary that the model be simulable subject to the conditioning information that is used to define the moment conditions. For this reason, the proposed estimator is applicable to general dynamic latent variable models. It is shown that as the number of simulations diverges, the estimator is consistent and a higher-order expansion reveals the stochastic difference between the infeasible GMM estimator based on the same moment conditions and the simulated version. In particular, we show how to adjust standard errors to account for the simulations. Monte Carlo results show how the estimator may be applied to a range of dynamic latent variable (DLV) models, and that it performs well in comparison to several other estimators that have been proposed for DLV models.

Nonparametric estimation of Value-at-Risk

A method to estimate an extreme quantile that requires no distributional assumptions is presented. The approach is based on transformed kernel estimation of the cumulative distribution function (cdf). The proposed method consists of a double transformation kernel estimation. We derive optimal bandwidth selection methods that have a direct expression for the smoothing parameter. The bandwidth can accommodate to the given quantile level. The procedure is useful for large data sets and improves quantile estimation compared to other methods in heavy tailed distributions. Implementation is straightforward and R programs are available.

Inequalities for a new data-based method for selecting nonparametric density estimates

We continue the development of a method for the selection of a bandwidth or a number of design parameters in density estimation. We provideexplicit non-asymptotic density-free inequalities that relate the $L_1$ error of the selected estimate with that of the best possible estimate,and study in particular the connection between the richness of the classof density estimates and the performance bound. For example, our methodallows one to pick the bandwidth and kernel order in the kernel estimatesimultaneously and still assure that for {\it all densities}, the $L_1$error of the corresponding kernel estimate is not larger than aboutthree times the error of the estimate with the optimal smoothing factor and kernel plus a constant times $\sqrt{\log n/n}$, where $n$ is the sample size, and the constant only depends on the complexity of the family of kernels used in the estimate. Further applications include multivariate kernel estimates, transformed kernel estimates, and variablekernel estimates.

Nonparametric Instrumental Regression

The focus of the paper is the nonparametric estimation of an instrumental regression function P defined by conditional moment restrictions stemming from a structural econometric model : E[Y-P(Z)|W]=0 and involving endogenous variables Y and Z and instruments W. The function P is the solution of an ill-posed inverse problem and we propose an estimation procedure based on Tikhonov regularization. The paper analyses identification and overidentification of this model and presents asymptotic properties of the estimated nonparametric instrumental regression function.

Extension of spherical nonparametric estimators to nonisotropic kernels: An oceanographic application

In a recently published paper. spherical nonparametric estimators were applied to feature-track ensembles to determine a range of statistics for the atmospheric features considered. This approach obviates the types of bias normally introduced with traditional estimators. New spherical isotropic kernels with local support were introduced. Ln this paper the extension to spherical nonisotropic kernels with local support is introduced, together with a means of obtaining the shape and smoothing parameters in an objective way. The usefulness of spherical nonparametric estimators based on nonisotropic kernels is demonstrated with an application to an oceanographic feature-track ensemble.

Spherical nonparametric estimators applied to the UGAMP model integration for AMIP

The aim of this paper is essentially twofold: first, to describe the use of spherical nonparametric estimators for determining statistical diagnostic fields from ensembles of feature tracks on a global domain, and second, to report the application of these techniques to data derived from a modern general circulation model. New spherical kernel functions are introduced that are more efficiently computed than the traditional exponential kernels. The data-driven techniques of cross-validation to determine the amount elf smoothing objectively, and adaptive smoothing to vary the smoothing locally, are also considered. Also introduced are techniques for combining seasonal statistical distributions to produce longer-term statistical distributions. Although all calculations are performed globally, only the results for the Northern Hemisphere winter (December, January, February) and Southern Hemisphere winter (June, July, August) cyclonic activity are presented, discussed, and compared with previous studies. Overall, results for the two hemispheric winters are in good agreement with previous studies, both for model-based studies and observational studies.

Nonparametric Message Passing Methods for Cooperative Localization and Tracking

The objective of this thesis is the development of cooperative localization and tracking algorithms using nonparametric message passing techniques. In contrast to the most well-known techniques, the goal is to estimate the posterior probability density function (PDF) of the position of each sensor. This problem can be solved using Bayesian approach, but it is intractable in general case. Nevertheless, the particle-based approximation (via nonparametric representation), and an appropriate factorization of the joint PDFs (using message passing methods), make Bayesian approach acceptable for inference in sensor networks. The well-known method for this problem, nonparametric belief propagation (NBP), can lead to inaccurate beliefs and possible non-convergence in loopy networks. Therefore, we propose four novel algorithms which alleviate these problems: nonparametric generalized belief propagation (NGBP) based on junction tree (NGBP-JT), NGBP based on pseudo-junction tree (NGBP-PJT), NBP based on spanning trees (NBP-ST), and uniformly-reweighted NBP (URW-NBP). We also extend NBP for cooperative localization in mobile networks. In contrast to the previous methods, we use an optional smoothing, provide a novel communication protocol, and increase the efficiency of the sampling techniques. Moreover, we propose novel algorithms for distributed tracking, in which the goal is to track the passive object which cannot locate itself. In particular, we develop distributed particle filtering (DPF) based on three asynchronous belief consensus (BC) algorithms: standard belief consensus (SBC), broadcast gossip (BG), and belief propagation (BP). Finally, the last part of this thesis includes the experimental analysis of some of the proposed algorithms, in which we found that the results based on real measurements are very similar with the results based on theoretical models.

Cooperative localization in mobile networks using nonparametric variants of belief propagation

Of the many state-of-the-art methods for cooperative localization in wireless sensor networks (WSN), only very few adapt well to mobile networks. The main problems of the well-known algorithms, based on nonparametric belief propagation (NBP), are the high communication cost and inefficient sampling techniques. Moreover, they either do not use smoothing or just apply it o ine. Therefore, in this article, we propose more flexible and effcient variants of NBP for cooperative localization in mobile networks. In particular, we provide: i) an optional 1-lag smoothing done almost in real-time, ii) a novel low-cost communication protocol based on package approximation and censoring, iii) higher robustness of the standard mixture importance sampling (MIS) technique, and iv) a higher amount of information in the importance densities by using the population Monte Carlo (PMC) approach, or an auxiliary variable. Through extensive simulations, we confirmed that all the proposed techniques outperform the standard NBP method.