Inférence robuste à la présence des valeurs aberrantes dans les enquêtes

Cette thèse comporte trois articles dont un est publié et deux en préparation. Le sujet central de la thèse porte sur le traitement des valeurs aberrantes représentatives dans deux aspects importants des enquêtes que sont : l’estimation des petits domaines et l’imputation en présence de non-réponse partielle. En ce qui concerne les petits domaines, les estimateurs robustes dans le cadre des modèles au niveau des unités ont été étudiés. Sinha & Rao (2009) proposent une version robuste du meilleur prédicteur linéaire sans biais empirique pour la moyenne des petits domaines. Leur estimateur robuste est de type «plugin», et à la lumière des travaux de Chambers (1986), cet estimateur peut être biaisé dans certaines situations. Chambers et al. (2014) proposent un estimateur corrigé du biais. En outre, un estimateur de l’erreur quadratique moyenne a été associé à ces estimateurs ponctuels. Sinha & Rao (2009) proposent une procédure bootstrap paramétrique pour estimer l’erreur quadratique moyenne. Des méthodes analytiques sont proposées dans Chambers et al. (2014). Cependant, leur validité théorique n’a pas été établie et leurs performances empiriques ne sont pas pleinement satisfaisantes. Ici, nous examinons deux nouvelles approches pour obtenir une version robuste du meilleur prédicteur linéaire sans biais empirique : la première est fondée sur les travaux de Chambers (1986), et la deuxième est basée sur le concept de biais conditionnel comme mesure de l’influence d’une unité de la population. Ces deux classes d’estimateurs robustes des petits domaines incluent également un terme de correction pour le biais. Cependant, ils utilisent tous les deux l’information disponible dans tous les domaines contrairement à celui de Chambers et al. (2014) qui utilise uniquement l’information disponible dans le domaine d’intérêt. Dans certaines situations, un biais non négligeable est possible pour l’estimateur de Sinha & Rao (2009), alors que les estimateurs proposés exhibent un faible biais pour un choix approprié de la fonction d’influence et de la constante de robustesse. Les simulations Monte Carlo sont effectuées, et les comparaisons sont faites entre les estimateurs proposés et ceux de Sinha & Rao (2009) et de Chambers et al. (2014). Les résultats montrent que les estimateurs de Sinha & Rao (2009) et de Chambers et al. (2014) peuvent avoir un biais important, alors que les estimateurs proposés ont une meilleure performance en termes de biais et d’erreur quadratique moyenne. En outre, nous proposons une nouvelle procédure bootstrap pour l’estimation de l’erreur quadratique moyenne des estimateurs robustes des petits domaines. Contrairement aux procédures existantes, nous montrons formellement la validité asymptotique de la méthode bootstrap proposée. Par ailleurs, la méthode proposée est semi-paramétrique, c’est-à-dire, elle n’est pas assujettie à une hypothèse sur les distributions des erreurs ou des effets aléatoires. Ainsi, elle est particulièrement attrayante et plus largement applicable. Nous examinons les performances de notre procédure bootstrap avec les simulations Monte Carlo. Les résultats montrent que notre procédure performe bien et surtout performe mieux que tous les compétiteurs étudiés. Une application de la méthode proposée est illustrée en analysant les données réelles contenant des valeurs aberrantes de Battese, Harter & Fuller (1988). S’agissant de l’imputation en présence de non-réponse partielle, certaines formes d’imputation simple ont été étudiées. L’imputation par la régression déterministe entre les classes, qui inclut l’imputation par le ratio et l’imputation par la moyenne sont souvent utilisées dans les enquêtes. Ces méthodes d’imputation peuvent conduire à des estimateurs imputés biaisés si le modèle d’imputation ou le modèle de non-réponse n’est pas correctement spécifié. Des estimateurs doublement robustes ont été développés dans les années récentes. Ces estimateurs sont sans biais si l’un au moins des modèles d’imputation ou de non-réponse est bien spécifié. Cependant, en présence des valeurs aberrantes, les estimateurs imputés doublement robustes peuvent être très instables. En utilisant le concept de biais conditionnel, nous proposons une version robuste aux valeurs aberrantes de l’estimateur doublement robuste. Les résultats des études par simulations montrent que l’estimateur proposé performe bien pour un choix approprié de la constante de robustesse.

Onset And Characterisation Of Chaos In A Two Dimensional Nonlinear Deterministic Model

Nature is full of phenomena which we call "chaotic", the weather being a prime example. What we mean by this is that we cannot predict it to any significant accuracy, either because the system is inherently complex, or because some of the governing factors are not deterministic. However, during recent years it has become clear that random behaviour can occur even in very simple systems with very few number of degrees of freedom, without any need for complexity or indeterminacy. The discovery that chaos can be generated even with the help of systems having completely deterministic rules - often models of natural phenomena - has stimulated a lo; of research interest recently. Not that this chaos has no underlying order, but it is of a subtle kind, that has taken a great deal of ingenuity to unravel. In the present thesis, the author introduce a new nonlinear model, a ‘modulated’ logistic map, and analyse it from the view point of ‘deterministic chaos‘.

On CD-systems of stateless deterministic R-automata with window size one

We study cooperating distributed systems (CD-systems) of restarting automata that are very restricted: they are deterministic, they cannot rewrite, but only delete symbols, they restart immediately after performing a delete operation, they are stateless, and they have a read/write window of size 1 only, that is, these are stateless deterministic R(1)-automata. We study the expressive power of these systems by relating the class of languages that they accept by mode =1 computations to other well-studied language classes, showing in particular that this class only contains semi-linear languages, and that it includes all rational trace languages. In addition, we investigate the closure and non-closure properties of this class of languages and some of its algorithmic properties.

CD-Systems of Stateless Deterministic R(1)-Automata Governed by an External Pushdown Store

We study cooperating distributed systems (CD-systems) of stateless deterministic restarting automata with window size 1 that are governed by an external pushdown store. In this way we obtain an automata-theoretical characterization for the class of context-free trace languages.

On Globally Deterministic CD-Systems of Stateless R-Automata with Window Size One

It is known that cooperating distributed systems (CD-systems) of stateless deterministic restarting automata with window size 1 accept a class of semi-linear languages that properly includes all rational trace languages. Although the component automata of such a CD-system are all deterministic, in general the CD-system itself is not, as in each of its computations, the initial component and the successor components are still chosen nondeterministically. Here we study CD-systems of stateless deterministic restarting automata with window size 1 that are themselves completely deterministic. In fact, we consider two such types of CD-systems, the strictly deterministic systems and the globally deterministic systems.

Enlarging a training set for genomic selction by imputation of un-genotyped animals in populations of varying genetic architecture

Background: The most common application of imputation is to infer genotypes of a high-density panel of markers on animals that are genotyped for a low-density panel. However, the increase in accuracy of genomic predictions resulting from an increase in the number of markers tends to reach a plateau beyond a certain density. Another application of imputation is to increase the size of the training set with un-genotyped animals. This strategy can be particularly successful when a set of closely related individuals are genotyped. ----- Methods: Imputation on completely un-genotyped dams was performed using known genotypes from the sire of each dam, one offspring and the offspring’s sire. Two methods were applied based on either allele or haplotype frequencies to infer genotypes at ambiguous loci. Results of these methods and of two available software packages were compared. Quality of imputation under different population structures was assessed. The impact of using imputed dams to enlarge training sets on the accuracy of genomic predictions was evaluated for different populations, heritabilities and sizes of training sets. ----- Results: Imputation accuracy ranged from 0.52 to 0.93 depending on the population structure and the method used. The method that used allele frequencies performed better than the method based on haplotype frequencies. Accuracy of imputation was higher for populations with higher levels of linkage disequilibrium and with larger proportions of markers with more extreme allele frequencies. Inclusion of imputed dams in the training set increased the accuracy of genomic predictions. Gains in accuracy ranged from close to zero to 37.14%, depending on the simulated scenario. Generally, the larger the accuracy already obtained with the genotyped training set, the lower the increase in accuracy achieved by adding imputed dams. ----- Conclusions: Whenever a reference population resembling the family configuration considered here is available, imputation can be used to achieve an extra increase in accuracy of genomic predictions by enlarging the training set with completely un-genotyped dams. This strategy was shown to be particularly useful for populations with lower levels of linkage disequilibrium, for genomic selection on traits with low heritability, and for species or breeds for which the size of the reference population is limited.

A deterministic model for lake clarity; application to management of Lake Tahoe (California-Nevada), USA

This dissertation has as its goal the quantitative evaluation of the application of coupled hydrodynamic, ecological and clarity models, to address the deterministic prediction of water clarity in lakes and reservoirs. Prediction of water clarity is somewhat unique, insofar as it represents the integrated and coupled effects of a broad range of individual water quality components. These include the biological components such as phytoplankton, together with the associated cycles of nutrients that are needed to sustain their popuiations, and abiotic components such as suspended particles that may be introduced by streams, atmospheric deposition or sediment resuspension. Changes in clarity induced by either component will feed back on the phytoplankton dynamics, as incident light also affects biological growth. Thus ability to successfully model changes in clarity will by necessity have to achieve the correct modeling of these other water quality parameters. Water clarity is also unique in that it may be one of the earliest and most easily detected wamings of the acceleration of the process of eutrophication in a water body.

Making development simple. The genetic deterministic hypothesis for economic development

This paper discusses the dangers inherent in allempting to simplify something as complex as development. It does this by exploring the Lynn and Vanhanen theory of deterministic development which asserts that varying levels of economic development seen between countries can be explained by differences in 'national intelligence' (national IQ). Assuming that intelligence is genetically determined, and as different races have been shown to have different IQ, then they argue that economic development (measured as GDP/capita) is largely a function of race and interventions to address imbalances can only have a limited impact. The paper presents the Lynne and Vanhanen case and critically discusses the data and analyses (linear regression) upon which it is based. It also extends the cause-effect basis of Lynne and Vanhanen's theory for economic development into human development by using the Human Development Index (HDI). It is argued that while there is nothing mathematically incorrect with their calculations, there are concerns over the data they employ. Even more fundamentally it is argued that statistically significant correlations between the various components of the HDI and national IQ can occur via a host of cause-effect pathways, and hence the genetic determinism theory is far from proven. The paper ends by discussing the dangers involved in the use of over-simplistic measures of development as a means of exploring cause-effect relationships. While the creators of development indices such as the HDI have good intentions, simplistic indices can encourage simplistic explanations of under-development. (c) 2005 Elsevier B.V. All rights reserved.

Adjusted jackknife for imputation under unequal probability sampling without replacement

Imputation is commonly used to compensate for item non-response in sample surveys. If we treat the imputed values as if they are true values, and then compute the variance estimates by using standard methods, such as the jackknife, we can seriously underestimate the true variances. We propose a modified jackknife variance estimator which is defined for any without-replacement unequal probability sampling design in the presence of imputation and non-negligible sampling fraction. Mean, ratio and random-imputation methods will be considered. The practical advantage of the method proposed is its breadth of applicability.

Performance and conformance results of the deterministic DVB-T equalizer

One of the enablers for new consumer electronics based products to be accepted in to the market is the availability of inexpensive, flexible and multi-standard chipsets and services. DVB-T, the principal standard for terrestrial broadcast of digital video in Europe, has been extremely successful in leading to governments reconsidering their targets for analogue television broadcast switch-off. To enable one further small step in creating increasingly cost effective chipsets, the ODFM deterministic equalizer has been presented before with its application to DVB-T. This paper discusses the test set-up of a DVB-T compliant baseband simulation that includes the deterministic equalizer and DVB-T standard propagation channels. This is then followed by a presentation of the found inner and outer Bit Error Rate (BER) results using various modulation levels, coding rates and propagation channels in order to ascertain the actual performance of the deterministic equalizer(1).

Deterministic equalization and results of a DVB-T multipath equalizer for both 16-QAM and 64-QAM operation

Recent developments in the UK concerning the reception of Digital Terrestrial Television (DTT) have indicated that, as it currently stands, DVB-T receivers may not be sufficient to maintain adequate quality of digital picture information to the consumer. There are many possible reasons why such large errors are being introduced into the system preventing reception failure. It has been suggested that one possibility is that the assumptions concerning the immunity to multipath that Coded Orthogonal Frequency Division Multiplex (COFDM) is expected to have, may not be entirely accurate. Previous research has shown that multipath can indeed have an impact on a DVB-T receiver performance. In the UK, proposals have been made to change the modulation from 64-QAM to 16-QAM to improve the immunity to multipath, but this paper demonstrates that the 16-QAM performance may again not be sufficient. To this end, this paper presents a deterministic approach to equalization such that a 64-QAM receiver with the simple equalizer presented in this paper has the same order of MPEG-2 BER performance as that to a 16-QAM receiver without equalization. Thus, alleviating the requirement in the broadcasters to migrate from 64-QAM to 16-QAM Of course, by adding the equalizer to a 16-QAM receiver then the BER is also further improved and thus creating one more step to satisfying the consumers(1).

Comparison of the computational cost of a Monte Carlo and deterministic algorithm for computing bilinear forms of matrix powers

In this paper we consider bilinear forms of matrix polynomials and show that these polynomials can be used to construct solutions for the problems of solving systems of linear algebraic equations, matrix inversion and finding extremal eigenvalues. An almost Optimal Monte Carlo (MAO) algorithm for computing bilinear forms of matrix polynomials is presented. Results for the computational costs of a balanced algorithm for computing the bilinear form of a matrix power is presented, i.e., an algorithm for which probability and systematic errors are of the same order, and this is compared with the computational cost for a corresponding deterministic method.

Deterministic IIR video deghoster for all ghost carrier phases

It is known that certain video deghoster systems cannot fully process the induced signal derived from the quadrature carrier forming nature of the VSB filter under a multipath condition. A new deterministic IIR deghoster filter structure is given which is capable of deghosting terrestrial video for any relative ghost carrier phase.

Working Group 1: Future developments in deterministic and stochastic parametrizations

Our group considered the desirability of including representations of uncertainty in the development of parameterizations. (By ‘uncertainty’ here we mean the deviation of sub-grid scale fluxes or tendencies in any given model grid box from truth.) We unanimously agreed that the ECWMF should attempt to provide a more physical basis for uncertainty estimates than the very effective but ad hoc methods being used at present. Our discussions identified several issues that will arise.