Parameter estimation in the presence of auxiliary information

Dissertação para obtenção do Grau de Doutora em Estatística e Gestão de Risco, Especialidade em Estatística

Design-based random permutation models with auxiliary information

We extend the random permutation model to obtain the best linear unbiased estimator of a finite population mean accounting for auxiliary variables under simple random sampling without replacement (SRS) or stratified SRS. The proposed method provides a systematic design-based justification for well-known results involving common estimators derived under minimal assumptions that do not require specification of a functional relationship between the response and the auxiliary variables.

Mapping of Environmental Data Using Kernel-Based Methods

Recently, kernel-based Machine Learning methods have gained great popularity in many data analysis and data mining fields: pattern recognition, biocomputing, speech and vision, engineering, remote sensing etc. The paper describes the use of kernel methods to approach the processing of large datasets from environmental monitoring networks. Several typical problems of the environmental sciences and their solutions provided by kernel-based methods are considered: classification of categorical data (soil type classification), mapping of environmental and pollution continuous information (pollution of soil by radionuclides), mapping with auxiliary information (climatic data from Aral Sea region). The promising developments, such as automatic emergency hot spot detection and monitoring network optimization are discussed as well.

Small-area estimation with spatial similarity

A class of composite estimators of small area quantities that exploit spatial (distancerelated)similarity is derived. It is based on a distribution-free model for the areas, but theestimators are aimed to have optimal design-based properties. Composition is applied alsoto estimate some of the global parameters on which the small area estimators depend.It is shown that the commonly adopted assumption of random effects is not necessaryfor exploiting the similarity of the districts (borrowing strength across the districts). Themethods are applied in the estimation of the mean household sizes and the proportions ofsingle-member households in the counties (comarcas) of Catalonia. The simplest version ofthe estimators is more efficient than the established alternatives, even though the extentof spatial similarity is quite modest.

Data-driven topo-climatic mapping with machine learning methods

Automatic environmental monitoring networks enforced by wireless communication technologies provide large and ever increasing volumes of data nowadays. The use of this information in natural hazard research is an important issue. Particularly useful for risk assessment and decision making are the spatial maps of hazard-related parameters produced from point observations and available auxiliary information. The purpose of this article is to present and explore the appropriate tools to process large amounts of available data and produce predictions at fine spatial scales. These are the algorithms of machine learning, which are aimed at non-parametric robust modelling of non-linear dependencies from empirical data. The computational efficiency of the data-driven methods allows producing the prediction maps in real time which makes them superior to physical models for the operational use in risk assessment and mitigation. Particularly, this situation encounters in spatial prediction of climatic variables (topo-climatic mapping). In complex topographies of the mountainous regions, the meteorological processes are highly influenced by the relief. The article shows how these relations, possibly regionalized and non-linear, can be modelled from data using the information from digital elevation models. The particular illustration of the developed methodology concerns the mapping of temperatures (including the situations of Föhn and temperature inversion) given the measurements taken from the Swiss meteorological monitoring network. The range of the methods used in the study includes data-driven feature selection, support vector algorithms and artificial neural networks.

Anonymisation, Deanonymisation and Data Management

Speaker: Dr Kieron O'Hara Organiser: Time: 04/02/2015 11:00-11:45 Location: B32/3077 Abstract In order to reap the potential societal benefits of big and broad data, it is essential to share and link personal data. However, privacy and data protection considerations mean that, to be shared, personal data must be anonymised, so that the data subject cannot be identified from the data. Anonymisation is therefore a vital tool for data sharing, but deanonymisation, or reidentification, is always possible given sufficient auxiliary information (and as the amount of data grows, both in terms of creation, and in terms of availability in the public domain, the probability of finding such auxiliary information grows). This creates issues for the management of anonymisation, which are exacerbated not only by uncertainties about the future, but also by misunderstandings about the process(es) of anonymisation. This talk discusses these issues in relation to privacy, risk management and security, reports on recent theoretical tools created by the UKAN network of statistics professionals (on which the author is one of the leads), and asks how long anonymisation can remain a useful tool, and what might replace it.

Road network and GPS tracking with data processing and quality assessment

GPS technology has been embedded into portable, low-cost electronic devices nowadays to track the movements of mobile objects. This implication has greatly impacted the transportation field by creating a novel and rich source of traffic data on the road network. Although the promise offered by GPS devices to overcome problems like underreporting, respondent fatigue, inaccuracies and other human errors in data collection is significant; the technology is still relatively new that it raises many issues for potential users. These issues tend to revolve around the following areas: reliability, data processing and the related application. This thesis aims to study the GPS tracking form the methodological, technical and practical aspects. It first evaluates the reliability of GPS based traffic data based on data from an experiment containing three different traffic modes (car, bike and bus) traveling along the road network. It then outline the general procedure for processing GPS tracking data and discuss related issues that are uncovered by using real-world GPS tracking data of 316 cars. Thirdly, it investigates the influence of road network density in finding optimal location for enhancing travel efficiency and decreasing travel cost. The results show that the geographical positioning is reliable. Velocity is slightly underestimated, whereas altitude measurements are unreliable.Post processing techniques with auxiliary information is found necessary and important when solving the inaccuracy of GPS data. The densities of the road network influence the finding of optimal locations. The influence will stabilize at a certain level and do not deteriorate when the node density is higher.

Validação de formulário de notificação de eventos adversos a plantas medicinais e fitoterápicos

Pós-graduação em Ciências Farmacêuticas - FCFAR

A new sampling strategy to reduce the effect of autocorrelation on a control chart

Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)

Estimación de variables dasométricas a partir de datos LiDAR y obtención de modelos de referencia para las distribuciones de alturas y diámetros del arbolado

La Gestión Forestal Sostenible se define como “la administración y uso de los bosques y tierras forestales de forma e intensidad tales que mantengan su biodiversidad, productividad, capacidad de regeneración, vitalidad y su potencial para atender, ahora y en el futuro, las funciones ecológicas, económicas y sociales relevantes a escala local, nacional y global, y que no causan daño a otros ecosistemas” (MCPFE Conference, 1993). Dentro del proceso los procesos de planificación, en cualquier escala, es necesario establecer cuál será la situación a la que se quiere llegar mediante la gestión. Igualmente, será necesario conocer la situación actual, pues marcará la situación de partida y condicionará el tipo de actuaciones a realizar para alcanzar los objetivos fijados. Dado que, los Proyectos de Ordenación de Montes y sus respectivas revisiones son herramientas de planificación, durante la redacción de los mismos, será necesario establecer una serie de objetivos cuya consecución pueda verificarse de forma objetiva y disponer de una caracterización de la masa forestal que permita conocer la situación de partida. Esta tesis se centra en problemas prácticos, propios de una escala de planificación local o de Proyecto de Ordenación de Montes. El primer objetivo de la tesis es determinar distribuciones diamétricas y de alturas de referencia para masas regulares por bosquetes, empleando para ello el modelo conceptual propuesto por García-Abril et al., (1999) y datos procedentes de las Tablas de producción de Rojo y Montero (1996). Las distribuciones de referencia obtenidas permitirán guiar la gestión de masas irregulares y regulares por bosquetes. Ambos tipos de masas aparecen como una alternativa deseable en aquellos casos en los que se quiere potenciar la biodiversidad, la estabilidad, la multifuncionalidad del bosque y/o como alternativa productiva, especialmente indicada para la producción de madera de calidad. El segundo objetivo de la Tesis está relacionado con la necesidad de disponer de una caracterización adecuada de la masa forestal durante la redacción de los Proyectos de Ordenación de Montes y de sus respectivas revisiones. Con el fin de obtener estimaciones de variables forestales en distintas unidades territoriales de potencial interés para la Ordenación de Montes, así como medidas de la incertidumbre en asociada dichas estimaciones, se extienden ciertos resultados de la literatura de Estimación en Áreas Pequeñas. Mediante un caso de estudio, se demuestra el potencial de aplicación de estas técnicas en inventario forestales asistidos con información auxiliar procedente de sensores láser aerotransportados (ALS). Los casos de estudio se realizan empleando datos ALS similares a los recopilados en el marco del Plan Nacional de Ortofotografía Aérea (PNOA). Los resultados obtenidos muestran que es posible aumentar la eficiencia de los inventarios forestales tradicionales a escala de proyecto de Ordenación de Montes, mediante la aplicación de estimadores EBLUP (Empirical Best Linear Unbiased Predictor) con modelos a nivel de elemento poblacional e información auxiliar ALS similar a la recopilada por el PNOA. ABSTRACT According to MCPFE (1993) Sustainable Forest Management is “the stewardship and use of forests and forest lands in a way, and at a rate, that maintains their biodiversity, productivity, regeneration capacity, vitality and their potential to fulfill, now and in the future, relevant ecological, economic and social functions, at local, national, and global levels, and that does not cause damage to other ecosystems”. For forest management planning, at any scale, we must determine what situation is hoped to be achieved through management. It is also necessary to know the current situation, as this will mark the starting point and condition the type of actions to be performed in order to meet the desired objectives. Forest management at a local scale is no exception. This Thesis focuses on typical problems of forest management planning at a local scale. The first objective of this Thesis is to determine management objectives for group shelterwood management systems in terms of tree height and tree diameter reference distributions. For this purpose, the conceptual model proposed by García-Abril et al., (1999) is applied to the yield tables for Pinus sylvestris in Sierra de Guadrrama (Rojo y Montero, 1996). The resulting reference distributions will act as a guide in the management of forests treated under the group shelterwood management systems or as an approximated reference for the management of uneven aged forests. Both types of management systems are desirable in those cases where forest biodiversity, stability and multifunctionality are pursued goals. These management systems are also recommended as alternatives for the production of high quality wood. The second objective focuses on the need to adequately characterize the forest during the decision process that leads to local management. In order to obtain estimates of forest variables for different management units of potential interest for forest planning, as well as the associated measures of uncertainty in these estimates, certain results from Small Area Estimation Literature are extended to accommodate for the need of estimates and reliability measures in very small subpopulations containing a reduced number of pixels. A case study shows the potential of Small Area Estimation (SAE) techniques in forest inventories assisted with remotely sensed auxiliary information. The influence of the laser pulse density in the quality of estimates in different aggregation levels is analyzed. This study considers low laser pulse densities (0.5 returns/m2) similar to, those provided by large-scale Airborne Laser Scanner (ALS) surveys, such as the one conducted by the Spanish National Geographic Institute for about 80% of the Spanish territory. The results obtained show that it is possible to improve the efficiency of traditional forest inventories at local scale using EBLUP (Empirical Best Linear Unbiased Predictor) estimators based on unit level models and low density ALS auxiliary information.

Polarimetric Response of Rice Fields at C-Band: Analysis and Phenology Retrieval

A set of ten RADARSAT-2 images acquired in fully polarimetric mode over a test site with rice fields in Seville, Spain, has been analyzed to extract the main features of the C-band radar backscatter as a function of rice phenology. After observing the evolutions versus phenology of different polarimetric observables and explaining their behavior in terms of scattering mechanisms present in the scene, a simple retrieval approach has been proposed. This algorithm is based on three polarimetric observables and provides estimates from a set of four relevant intervals of phenological stages. The validation against ground data, carried out at parcel level for a set of six stands and up to nine dates per stand, provides a 96% rate of coincidence. Moreover, an equivalent compact-pol retrieval algorithm has been also proposed and validated, providing the same performance at parcel level. In all cases, the inversion is carried out by exploiting a single satellite acquisition, without any other auxiliary information.

Collaborative filtering and deep learning based recommendation system for cold start items

Recommender system is a specific type of intelligent systems, which exploits historical user ratings on items and/or auxiliary information to make recommendations on items to the users. It plays a critical role in a wide range of online shopping, e-commercial services and social networking applications. Collaborative filtering (CF) is the most popular approaches used for recommender systems, but it suffers from complete cold start (CCS) problem where no rating record are available and incomplete cold start (ICS) problem where only a small number of rating records are available for some new items or users in the system. In this paper, we propose two recommendation models to solve the CCS and ICS problems for new items, which are based on a framework of tightly coupled CF approach and deep learning neural network. A specific deep neural network SADE is used to extract the content features of the items. The state of the art CF model, timeSVD++, which models and utilizes temporal dynamics of user preferences and item features, is modified to take the content features into prediction of ratings for cold start items. Extensive experiments on a large Netflix rating dataset of movies are performed, which show that our proposed recommendation models largely outperform the baseline models for rating prediction of cold start items. The two proposed recommendation models are also evaluated and compared on ICS items, and a flexible scheme of model retraining and switching is proposed to deal with the transition of items from cold start to non-cold start status. The experiment results on Netflix movie recommendation show the tight coupling of CF approach and deep learning neural network is feasible and very effective for cold start item recommendation. The design is general and can be applied to many other recommender systems for online shopping and social networking applications. The solution of cold start item problem can largely improve user experience and trust of recommender systems, and effectively promote cold start items.

Histopathological examination of nerve samples from pure neural leprosy patients: obtaining maximum information to improve diagnostic efficiency

Nerve biopsy examination is an important auxiliary procedure for diagnosing pure neural leprosy (PNL). When acid-fast bacilli (AFB) are not detected in the nerve sample, the value of other nonspecific histological alterations should be considered along with pertinent clinical, electroneuromyographical and laboratory data (the detection of Mycobacterium leprae DNA with polymerase chain reaction and the detection of serum anti-phenolic glycolipid 1 antibodies) to support a possible or probable PNL diagnosis. Three hundred forty nerve samples [144 from PNL patients and 196 from patients with non-leprosy peripheral neuropathies (NLN)] were examined. Both AFB-negative and AFB-positive PNL samples had more frequent histopathological alterations (epithelioid granulomas, mononuclear infiltrates, fibrosis, perineurial and subperineurial oedema and decreased numbers of myelinated fibres) than the NLN group. Multivariate analysis revealed that independently, mononuclear infiltrate and perineurial fibrosis were more common in the PNL group and were able to correctly classify AFB-negative PNL samples. These results indicate that even in the absence of AFB, these histopathological nerve alterations may justify a PNL diagnosis when observed in conjunction with pertinent clinical, epidemiological and laboratory data.

Optimal information transmission in organizations: Search and congestion

We propose a stylized model of a problem-solving organization whoseinternal communication structure is given by a fixed network. Problemsarrive randomly anywhere in this network and must find their way to theirrespective specialized solvers by relying on local information alone.The organization handles multiple problems simultaneously. For this reason,the process may be subject to congestion. We provide a characterization ofthe threshold of collapse of the network and of the stock of foatingproblems (or average delay) that prevails below that threshold. We buildupon this characterization to address a design problem: the determinationof what kind of network architecture optimizes performance for any givenproblem arrival rate. We conclude that, for low arrival rates, the optimalnetwork is very polarized (i.e. star-like or centralized ), whereas it islargely homogenous (or decentralized ) for high arrival rates. We also showthat, if an auxiliary assumption holds, the transition between these twoopposite structures is sharp and they are the only ones to ever qualify asoptimal.

The pernicious cycle of vlus in brazil: Epidemiology, pathogeny and auxiliary healing methods

Venous leg ulcers (VLUs) represent the most advanced stage of chronic venous insufficiency. Despite the large body of knowledge available regarding the risk factors and aetiopathogeny of the condition, patients referred to public health care systems in developing countries often do not receive adequate diagnosis or early treatment, leading to clinical evolution and disease recurrence. This review collates updated information about the epidemiology, risk factors, aetiopathogeny, diagnosis, ulcer healing methods and determinant factors of the pernicious cycle of VLUs in developing countries, with a focus on the Brazilian setting.