published_or_final_version / Statistics and Actuarial Science / Doctoral / Doctor of Philosophy

published_or_final_version / abstract / Social Sciences / Doctoral / Doctor of Philosophy

When justice goes wrong, grave consequences entail. They are damaging to the standing of the legal system and people’s lives. Humans are not good at assessing uncertainties. Parties to a legal proceeding adduce evidence to support or reject hypotheses. Errors happen when the tribunal fails to consider properly all the evidence in the context of inherent probabilities or improbabilities. This research work does not advocate trials by mathematics or statistics. This work extends the understanding of the application of Bayesian Networks in the law domain. The original contribution to knowledge is the analysis of evidence by Bayesian Network in the context of specific legal requirements of Hong Kong. The research questions are: 1. What are the legal requirements for the analysis of evidence in a criminal trial in Hong Kong? 2. How can a Bayesian Network be constructed for the purpose of such analysis? 3. Is such a Bayesian Network effective for the analysis? In answering the questions, this research work examined the feasibility of generic models created for digital crime scene investigations and concluded that each case must be, for the purpose of analysis of evidence in the trial, represented by a different Bayesian Network. This research work examined the trial processes, the tasks of tribunal of facts of criminal trials and some appellate decisions in Hong Kong. The work also created models of reasoning processes for the juries in Hong Kong. The work then compared the properties of Bayesian Networks with the processes of evaluation of evidence during trials. This research work also considered the reluctance of courts in the United Kingdom to allow experts to express their opinions on the bases of Bayesian calculations; even though trials are practically evaluations of uncertainties and assignments of degrees of beliefs. This research work then constructed a schedule of levels of proof and proposed a schematic method, to be used with the schedule, to construct Bayesian Networks for any types of trial in Hong Kong. The method requires an analyst to go through a mass of evidence systematically, and analyse their relationships amongst the ultimate probandum, the penultimate probanda, the intermediate propositions, the facts in issue and the facts relevant to the issue. This work then demonstrated the applications by two criminal cases in Hong Kong. The analyses show that the construction of Bayesian Network by the schematic method enables an analyst to take precaution to reach an assessment rationally and to approximate as far as capable his or her belief to the facts in issue. / published_or_final_version / Computer Science / Doctoral / Doctor of Philosophy

The accident rate at an intersection is one attribute used to evaluate the hazard associated with the intersection. Two techniques traditionally used to make such evaluations are the rate-quality technique and a technique based on the confidence interval of classical statistics. Both of these techniques label intersections as hazardous if their accident rate is greater than some critical accident rate determined by the technique. An alternative technique is one based on a Bayesian analysis of available accident number and traffic volume data. In contrast to the two classic techniques, the Bayesian technique identifies an intersection as hazardous based on a probabilistic assessment of accident rates. The goal of this thesis is to test and compare the ability of the three techniques to accurately identify traffic intersections known to be hazardous. Test data is generated from an empirical distribution of accident rates. The techniques are then applied to the generated data and compared based on the simulation results.

It is well-known that the observation of a variable in a Bayesian network can affect the effective connectivity of the network, which in turn affects the efficiency of inference. Unfortunately, the observed variables may not be known until runtime, which limits the amount of compile-time optimization that can be done in this regard. This thesis considers how to improve inference when users know the likelihood of a variable being observed. It demonstrates how these probabilities of observation can be exploited to improve existing heuristics for choosing elimination orderings for inference. Empirical tests over a set of benchmark networks using the Variable Elimination algorithm show reductions of up to 50% and 70% in multiplications and summations, as well as runtime reductions of up to 55%. Similarly, tests using the Elimination Tree algorithm show reductions by as much as 64%, 55%, and 50% in recursive calls, total cache size, and runtime, respectively. / xi, 88 leaves : ill. ; 29 cm

In many settings, agents engage in problem-solving activities, which require them to share resources, act on each others behalf, coordinate individual acts, etc. If autonomous agents are to e ectively interact (or support interaction among humans) in situations such as deciding whom and how to approach the provision of a resource or the performance of an action, there are a number of important questions to address. Who do I choose to delegate a task to? What do I need to say to convince him/her to do something? Were similar requests granted from similar agents in similar circumstances? What arguments were most persuasive? What are the costs involved in putting certain arguments forward? Research in argumentation strategies has received signi cant attention in recent years, and a number of approaches has been proposed to enable agents to reason about arguments to present in order to persuade another. However, current approaches do not adequately address situations where agents may be operating under social constraints (e.g., policies) that regulate behaviour in a society. In this thesis, we propose a novel combination of techniques that takes into consideration the policies that others may be operating with. First, we present an approach where evidence derived from dialogue is utilised to learn the policies of others. We show that this approach enables agents to build more accurate and stable models of others more rapidly. Secondly, we present an agent decision-making mechanism where models of others are used to guide future argumentation strategy. This approach takes into account the learned policy constraints of others, the cost of revealing in- formation, and anticipated resource availability in deciding whom to approach. We empirically evaluate our approach within a simulated multi-agent frame- work, and demonstrate that through the use of informed strategies agents can improve their performance.

Rural inhabitants of developing countries face extraordinarily risky environments, and decision-making under risk has crucial implications for the welfare of the rural poor. Therefore, obtaining a better understanding of the behaviour under risk of low-income populations is a vital step in the comprehension of human behaviour, and is important for effective policy design and evaluation, as well as for shedding light on production, investment and technology adoption decisions. In Chapter One, I analyze data collected from a laboratory experiment involving poor subjects in rural Ethiopia, in order to determine which decision models (and corresponding risk preferences) best describe the decision-making under risk of inhabitants. I find that expected utility theory (EUT) does not provide a good overall description of the decisions made by participants in the experiment; instead, there is evidence of probability weighting and loss aversion, implying that rank-dependent and reference-dependent choice models are more likely to represent the true latent decision-making process of subjects. In Chapter Two, I analyze combined experimental and survey data from rural Ethiopia in order to evaluate the determinants of risk preferences as well as assess the degree of asset integration in experimental decisions. Analyzing both EUT and non-EUT decision models and using an instrumental variable strategy, I find that household wealth negatively affects both risk aversion and loss aversion, but independent background risk has no effect on risk preferences. Further, I find evidence of narrow framing, as opposed to asset integration, suggesting that participants make decisions in the experiment in isolation from outside wealth. In Chapter Three, I analyze experimental data from Brazil to evaluate whether subjects understand decision problems that use the complex Multiple Price List (MPL) elicitation procedure, and to determine which decision models best describe observed choices. I find that the MPL decision problems of the experiment enable a finer characterization of risk preferences as compared to Ordered Lottery Selection problems (used in the Ethiopian experiment). However, I find that a significant fraction of choice patterns in the MPL problems are intransitive, and the evidence indicates that subjects did not properly understand the decision problems and thus observed choices do not reveal true risk preferences. Therefore, the relatively complex MPL procedure may not be suitable for experiments conducted with poorly-educated subjects in developing country settings. Chapter Four presents a theory outlining the relationship between rational demand for index insurance – for which the net transfer between insurer and policyholders depends only on a publicly verifiable index – and wealth. Further, the validity of this theory is tested using the experimental data from Ethiopia. In line with the theoretical model presented, due to basis risk and actuarially unfair premiums, demand for index insurance is hump-shaped – first increasing then decreasing – in wealth. The results indicate that the low take-up of this product observed among the poorest (and most risk averse) individuals in recent field studies may result from rational choice rather than credit constraints or poor decision-making.

A research report submitted to the Faculty of Engineering and the Built Environment, University of the Witwatersrand, Johannesburg, in fulfilment of the requirements for the degree of Master of Science in Engineering, 2017 / The main problem to be investigated is how short duration data records can be augmented using existing data from nearby catchments with data with long periods of record. The purpose of the investigation is to establish a method of improving hydrological data using data from a gauged catchment to improve data from an ungauged catchment. The investigation is undertaken using rainfall data for catchments in Lesotho. Marginal distributions describing the annual maximum rainfall for the catchments, and a joint distribution of pairs of catchments were established. The parameters of these distributions were estimated using the Bayesian – Markov Chain Monte Carlo approach, and using both the single-site (univariate) estimation and the two-site (bivariate) estimations. The results of the analyses show that for catchments with data with short periods of record, the precision of the estimated location and scale parameters improved when the estimates were carried out using the two-site (bivariate) method. Rainfall events predicted using bivariate analyses parameters were generally higher than the univariate analyses parameters. From the results, it can be concluded that the two-site approach can be used to improve the precision of the rainfall predictions for catchments with data with short periods of record. This method can be used in practice by hydrologists and design engineers to enhance available data for use in designs and assessments. / CK2018

Diversos aspectos subjetivos influenciam o cliente no momento da aquisição de um produto. Estética, qualidade percebida e conforto são alguns deles, se o produto não atender às expectativas do cliente relacionadas a isto, corre o risco de se tornar um fracasso de vendas. Os objetivos do produto, quanto aos aspectos subjetivos, denominam-se requisitos subjetivos e são estabelecidos na primeira etapa do desenvolvimento de um novo produto (NPD). Tratá-los como imutáveis ao longo do NPD não condiz com a realidade atual de velozes mudanças. A literatura ainda é inexpressiva em relação à gestão das mudanças dos requisitos subjetivos. Este trabalho busca mostrar a importância de gerenciá-los ao longo de todo o NPD. Quanto mais longo for o ciclo desse processo, maior a chance de que os requisitos precisem ser revistos. Este estudo de caso, do interior do jato executivo Legacy 500, mostra algumas das consequências resultantes da falta de um processo para gerenciar os requisitos subjetivos. Os possíveis impactos estão relacionados com atrasos tanto no próprio projeto do produto em questão, como nos produtos que seriam desenvolvidos na sequência com o mesmo corpo de engenheiros, projetistas e designers e o aumento considerável de custos. Os resultados desta pesquisa apontam fatores internos e externos que podem afetar os requisitos subjetivos e sua gestão. Os internos estão relacionados à estratégia e ao posicionamento da marca, ao processo de gestão, à estrutura organizacional e à competência do(s) decisor(es) em design de produto, além de um profundo conhecimento do mercado e de quem é o potencial cliente, tanto por parte do decisor estratégico, quanto do time operacional. Já os externos estão relacionados às mudanças no mercado, aos novos produtos, às novas tecnologias e aos novos desejos/necessidades dos clientes. Diferentes sistemas, podem sofrer influência destes fatores, de distintas maneiras, no caso de um produto complexo. Também se dedica atenção ao processo decisório relacionado aos requisitos subjetivos, colaborando com a teoria e com a prática desta gestão, sugerindo um modelo que deixa evidente a importância do veículo de análise utilizado e das características do decisor. A natureza subjetiva de muitos aspectos do design do produto parecem conduzir a uma decisão do tipo intuitiva, diretamente relacionada com a experiência formal e profissional do(s) decisor(es). A representatividade do veículo de análise utilizado na avaliação dos requisitos subjetivos precisa ser a maior possível em todos os seus aspectos, já que o ser humano utiliza de todos os seus sentidos para formar a percepção que tem sobre um produto. O protótipo ou outro veículo de análise deve estar completo. A avaliação de partes em separado não gera a mesma percepção que sua avaliação como um todo. Materiais de acabamento, volumetria, resistência e rigidez de mecanismos e de componentes, interfaces, funcionalidades, temperatura, iluminação, ruídos e músicas são alguns dos exemplos que interferem na formação da percepção correta sobre o produto. Um produto que não é atraente para o cliente, não é lucrativo para a empresa. Por isso, gerenciar requisitos subjetivos é vital para a saúde do negócio. / Several subjective aspects influence the customer when purchasing a product. Aesthetics, perceived quality and comfort are some of them, if the product does not meet customer expectations related to this, it runs the risk of becoming a sales failure. The objectives of the product for the subjective aspects are called subjective requirements, they are established in the first stage of the development of a new product (NPD). Treating them as immutable throughout the NPD does not match the current reality of swift changes. The literature is still inexpressive in relation to the management of changes in subjective requirements. This work seeks to show the importance of managing them throughout the entire NPD. The longer the cycle of this process, the greater the chance that requirements need to be revised. This case study from the interior of the Legacy 500 executive jet shows some of the consequences resulting from the lack of a process to manage subjective requirements. The possible impacts, if perceived to be a need for change only during the final evaluation, already on the eve of entering the product on the market, are related to delays both in the project of the product itself and in the products that would be developed in sequence with the same body of engineers, designers and industrial designers and the considerable increase of costs. The results of this research point to internal and external factors that may affect the subjective requirements and their management. Internal factors are related to the strategy and positioning of the brand, the management process, the organizational structure and the competence of the decision maker (s) in product design, as well as a deep knowledge of the market and of who is the potential customer, that is need from both the strategic decision maker and the operational team. The external ones are related to changes in the market, new products, new technologies and new wishes / needs of the clients. Different systems may or may not be influenced from these factors in different ways in the case of a complex product. Attention is also given to the decision-making process related to subjective requirements, collaborating with the theory and practice of this management, suggesting a model that makes clear the importance of the analysis vehicle used and the characteristics of the decision maker. The subjective nature of many aspects of product design seem to lead to an intuitive decision, directly related to the formal and professional experience of the decision maker (s). The representativeness of the vehicle of analysis used in the evaluation of subjective requirements must be the greatest possible in all its aspects, since the human being uses all of his senses to form the perception he has about the product. The prototype or other analysis vehicle must be complete. The evaluation of separate parts does not generate the same perception as the complete set. Finishing materials, volumetry, resistance and rigidity of mechanisms and components, interfaces, functionalities, temperature, lighting, noise, music and other sounds are some of the examples that interfere in the formation of the correct perception about the product. A product that is not attractive to the customer, is not profitable for the company. Therefore, managing subjective requirements is vital to the business health

Keywords: Latent variables, Ordered categorical data, Unordered categorical data, Nonignorable missing data, Maximum likelihood approach, Bayesian approach. / Structural equation models (SEMs) have been widely applied to examine interrelationships among latent and observed variables in behavioral, psychological, and medical research. Motivated by the fact that correlated ordered and unordered categorical variables are frequently encountered in practical applications, a nonlinear SEM that accommodates fixed covariates, mixed continuous, ordered categorical, and unordered categorical variables is proposed. Maximum likelihood methods for estimation and model comparison are discussed. Besides, missing data are frequently encountered in practical researches; a lot of attention has been devoted to analyze various SEMs with missing data. Bayesian analysis, including parameter estimate and model comparison, of a nonlinear SEM with mixed continuous, ordered and unordered categorical variables, and nonignorable missing entries is also considered in the thesis. Simulation studies are conducted to reveal the performance of the proposed methods. Moreover, we apply our methodologies to analyze the real-life data set about cardiovascular disease. As none of the existing SEMs can simultaneously accommodate fixed covariates, mixed continuous, ordered and unordered categorical data, and missing data, this thesis offers a novel SEM to cope with more complex practical problems and develop maximum likelihood and Bayesian methods for obtaining results in estimation and model comparison. / Cai, Jingheng. / Advisers: Sik-Yum Lee; Xin-Yuan Song. / Source: Dissertation Abstracts International, Volume: 70-06, Section: B, page: 3584. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2008. / Includes bibliographical references (leaves 76-82). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [200-] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstracts in English and Chinese. / School code: 1307.