Neste trabalho estudamos o modelo de regressão logística com erro de medida nas covariáveis. Abordamos as metodologias de estimação de máxima pseudoverossimilhança pelo algoritmo EM-Monte Carlo, calibração da regressão, SIMEX e naïve (ingênuo), método este que ignora o erro de medida. Comparamos os métodos em relação à estimação, através do viés e da raiz do erro quadrático médio, e em relação à predição de novas observações, através das medidas de desempenho sensibilidade, especificidade, verdadeiro preditivo positivo, verdadeiro preditivo negativo, acurácia e estatística de Kolmogorov-Smirnov. Os estudos de simulação evidenciam o melhor desempenho do método de máxima pseudoverossimilhança na estimação. Para as medidas de desempenho na predição não há diferença entre os métodos de estimação. Por fim, utilizamos nossos resultados em dois conjuntos de dados reais de diferentes áreas: área médica, cujo objetivo está na estimação da razão de chances, e área financeira, cujo intuito é a predição de novas observações. / We study the logistic model when explanatory variables are measured with error. Three estimation methods are presented, namely maximum pseudo-likelihood obtained through a Monte Carlo expectation-maximization type algorithm, regression calibration, SIMEX and naïve, which ignores the measurement error. These methods are compared through simulation. From the estimation point of view, we compare the different methods by evaluating their biases and root mean square errors. The predictive quality of the methods is evaluated based on sensitivity, specificity, positive and negative predictive values, accuracy and the Kolmogorov-Smirnov statistic. The simulation studies show that the best performing method is the maximum pseudo-likelihood method when the objective is to estimate the parameters. There is no difference among the estimation methods for predictive purposes. The results are illustrated in two real data sets from different application areas: medical area, whose goal is the estimation of the odds ratio, and financial area, whose goal is the prediction of new observations.

Made available in DSpace on 2015-12-03T18:33:13Z (GMT). No. of bitstreams: 1 Eudes_Martins_de_Araujo.pdf: 471826 bytes, checksum: 3fa94e3e0d87a5b29960d321fc0f5277 (MD5) Previous issue date: 2015-03-31 / The objective of this work is to verify if the credit rating model proposed by Brito and Assaf Neto (2008) designed for publicly held companies may also be applied to privately held companies in Brazil. In this work, 60 companies were used, being 30 of them in bankruptcy or insolvency processes in the period from 1994 to 2004, herein referred to as insolvent companies, and 30 of them with normal economic and financial situation referred here as solvent companies. In the present study, 60 companies were used; 30 of them presenting financial restrictions during the year of 2013 and 30 having a normal economic and financial situation. The model proposed by Brito and Assaf Neto (2008) used a logistic regression with 25 economic and financial indicators to see if they were able to separate solvent companies from non-solvent companies. Out of the 25 indicators used for this study, only 4 of them were statistically significant, namely: (I) retained profits on assets, (ii) financial debt, (III) net working capital and (IV) cash balance on sales. This four-variable model obtained a 90% accuracy in the correct classification of solvent and insolvent companies. However, the logistic regression model estimated based on the data from private companies showed different results from the one estimated by Brito and Assaf Neto (2008).In this case, only two variables showed to be statistically significant: (I) equity on assets and (II) cash balance on sales. This adjusted model reached a 57% accuracy in correctly classifying the companies. In short, the results presented here showed that it was not possible to estimate the adjusted credit-scoring model with a good accuracy for privately held companies in Brazil this based on extracted data from their financial statements. / O objetivo neste trabalho ?? verificar se o modelo de classifica????o de risco de cr??dito proposto por Brito e Assaf Neto (2008) desenvolvido para companhias de capital aberto tamb??m pode ser aplicado as companhias de capital fechado no Brasil. Nele foram utilizadas 60 companhias, sendo 30 com processos de concordata ou fal??ncia no per??odo de 1994 a 2004 denominadas insolventes e 30 com situa????o econ??mico-financeira normal denominadas solventes. No estudo aqui desenvolvido, tamb??m foram utilizadas 60 companhias; 30 apresentando restritivos financeiros durante o ano de 2013 e 30 com situa????o econ??mico-financeira normal. O modelo proposto por Brito e Assaf Neto (2008) utilizou uma regress??o log??stica com 25 indicadores econ??mico-financeiros para verificar se eles eram capazes de separar companhias solventes de companhias insolventes. Dos 25 indicadores utilizados, apenas 4 deles apresentaram signific??ncia estat??stica, sendo eles: (I) lucros retidos sobre ativo, (II) endividamento financeiro, (III) capital de giro l??quido e (IV) saldo de tesouraria sobre vendas. Esse modelo com quatro vari??veis obteve uma acur??cia de 90% nas classifica????es corretas das companhias abertas solventes e insolventes. No entanto, o modelo de regress??o log??stica estimado com base nos dados das companhias de capital fechado mostrou resultados distintos daquele estimado por Brito e Assaf Neto (2008). Nesse caso, apenas duas vari??veis se mostraram estatisticamente significantes: (I) patrim??nio l??quido sobre ativo e (II) saldo de tesouraria sobre vendas. Esse modelo ajustado obteve uma acur??cia de apenas 57% nas classifica????es corretas das companhias. Em suma os resultados aqui relatados mostraram que n??o foi poss??vel estimar um modelo ajustado de credit scoring com boa acur??cia para companhias de capital fechado no Brasil com base em dados extra??dos de suas demonstra????es financeiras.

With the advent of computers and the information age, vast amounts of data generated in a great deal of science and industry fields require the statisticians to explore further. In particular, statistical and computational problems in biology and medicine have created a new field of bioinformatics, which is attracting more and more statisticians, computer scientists, and biologists. Several procedures have been developed for tracing the source of fecal pollution in water resources based on certain characteristics of certain microorganisms. Use of this collection of techniques has been termed microbial source tracking (MST). Most of the current methods for MST are based on patterns of either phenotypic or genotypic variation in indicator organisms. Studies also suggested that patterns of genotypic variation might be more reliable due to their less association with environmental factors than those of phenotypic variation. Among the genotypic methods for source tracking, fingerprinting via rep-PCR is most common. Thus, identifying the specific pollution sources in contaminated waters based on rep-PCR fingerprinting techniques, viewed as a classification problem, has become an increasingly popular research topic in bioinformatics. In the project, several statistical methods for classification were studied, including linear discriminant analysis, quadratic discriminant analysis, logistic regression, and $k$-nearest-neighbor rules, neural networks and support vector machine. This project report summaries each of these methods and relevant statistical theory. In addition, an application of these methods to a particular set of MST data is presented and comparisons are made.

Predictive power of classification models can be evaluated by various measures. The most popular measures in data mining (DM) are Gini coefficient, Kolmogorov-Smirnov statistic and lift. These measures are each based on a completely different way of calculation. If an analyst is used to one of these measures it can be difficult for him to asses the predictive power of a model evaluated by another measure. The aim of this thesis is to develop a method how to convert one performance measure into another. Even though this thesis focuses mainly on the above-mentioned measures, it deals also with other measures like sensitivity, specificity, total accuracy and area under ROC curve. During development of DM models you may need to work with a sample that is stratified by values of the target variable Y instead of working with the whole population containing millions of observations. If you evaluate a model developed on a stratified data you may need to convert these measures to the whole population. This thesis describes a way, how to carry out this conversion. A software application (CPM) enabling all these conversions makes part of this thesis. With this application you can not only convert one performance measure to another, but you can also convert measures calculated on a stratified sample to the whole population. Besides the above mentioned performance measures (sensitivity, specificity, total accuracy, Gini coefficient, Kolmogorov-Smirnov statistic), CPM will also generate confusion matrix and performance charts (lift chart, gains chart, ROC chart and KS chart). This thesis comprises the user manual to this application as well as the web address where the application can be downloaded. The theory described in this thesis was verified on the real data.

L’étude du rôle de l’activisme actionnarial des hedge funds (HF) en matière de gouvernancedans la création de valeur des entreprises françaises constitue l’objet principal de cette thèse.Les réflexions menées, issues de la théorie de la gouvernance nous ont permis d’apporter lespremiers éléments de réponse à notre question centrale. L’approche exploratoire de cephénomène nous a conduits à déterminer un modèle français de l’activisme des HF. De notreapproche hypothético-déductive, nous avons pu élaborer un profil type des entreprises viséescotées et examiner les déterminants de la réaction du marché financier avant de mesurerl’effet de l’activisme des HF sur la création de valeur. Il en ressort que ce dernier est àl’origine de nouvelles situations d’équilibre organisationnel. Il est toutefois entravé par laréticence de la coalition de contrôle, renforcée par les caractéristiques spécifiques de lagouvernance et du contexte juridique en France, à entreprendre ces aménagements. Résultat,des coûts de conviction excessifs, des coûts d’enracinement et l’attentisme des investisseurssur le marché français. Contrairement aux investisseurs traditionnels, il est à admettre quel’activisme des HF est un mécanisme alternatif de contrôle efficient, là où certainsmécanismes de gouvernance actionnariale et partenariale semblent être insuffisammentdisciplinaires. Cependant, il est insuffisant pour inciter la coalition de contrôle à agir dansl’intérêt des différentes parties prenantes de l’entreprise. En France, la concentration et lanature actionnariale limitent, voire bloquent, les actions activistes des HF, contrairement auxentreprises américaines où le capital n’est pas verrouillé. La performance des entreprisesfrançaises est donc l’apanage du pouvoir managérial face à une influence activiste des HF. / This research aims to analyse the role of hedge fund (HF) activism in corporate governanceand French firms value creation. The reflections devoted the theory of governance haveenabled us to provide first elements to our central question. The exploratory study allowed usto determine a French model of HF activism. Our hypothetical-deductive approach enabledus to develop a typical profile of a target-listed firm and to examine the determinants of themarket reaction before measuring the effect of HF activism on value creation. The resultsshow that HF activism is causing new organizational equilibrium situations. However, it isopposed by the reluctance of the control coalition due to French governance and legalcontext. Hence, the exorbitant costs of activism in addition to entrenchment problems andinvestors wait-and-see attitude on the French market. As opposed to traditional investors, HFactivism appears as an efficient alternative control mechanism, where some shareholder andstakeholder governance mechanisms seem to be insufficiently disciplinary. Still, HF activismis unable to encourage the reluctant control coalition to act in the firms’ stakeholders’interest. In France, the controlling and family shareholders are opposed to HF activism. It ismore difficult for them to act as they do in American firms where capital is not locked. Thevalue creation is thus confined to managerial power over HF activism.

Neste trabalho estudamos o modelo de regressão logística com erro de medida nas covariáveis. Abordamos as metodologias de estimação de máxima pseudoverossimilhança pelo algoritmo EM-Monte Carlo, calibração da regressão, SIMEX e naïve (ingênuo), método este que ignora o erro de medida. Comparamos os métodos em relação à estimação, através do viés e da raiz do erro quadrático médio, e em relação à predição de novas observações, através das medidas de desempenho sensibilidade, especificidade, verdadeiro preditivo positivo, verdadeiro preditivo negativo, acurácia e estatística de Kolmogorov-Smirnov. Os estudos de simulação evidenciam o melhor desempenho do método de máxima pseudoverossimilhança na estimação. Para as medidas de desempenho na predição não há diferença entre os métodos de estimação. Por fim, utilizamos nossos resultados em dois conjuntos de dados reais de diferentes áreas: área médica, cujo objetivo está na estimação da razão de chances, e área financeira, cujo intuito é a predição de novas observações. / We study the logistic model when explanatory variables are measured with error. Three estimation methods are presented, namely maximum pseudo-likelihood obtained through a Monte Carlo expectation-maximization type algorithm, regression calibration, SIMEX and naïve, which ignores the measurement error. These methods are compared through simulation. From the estimation point of view, we compare the different methods by evaluating their biases and root mean square errors. The predictive quality of the methods is evaluated based on sensitivity, specificity, positive and negative predictive values, accuracy and the Kolmogorov-Smirnov statistic. The simulation studies show that the best performing method is the maximum pseudo-likelihood method when the objective is to estimate the parameters. There is no difference among the estimation methods for predictive purposes. The results are illustrated in two real data sets from different application areas: medical area, whose goal is the estimation of the odds ratio, and financial area, whose goal is the prediction of new observations.

The aim of this thesis was to investigate the use of effect sizes to report the results of statistical credit rating models in a more practical way. Rating systems in the form of statistical probability models like logistic regression models are used to forecast the behaviour of clients and guide business in rating clients as “high” or “low” risk borrowers. Therefore, model results were reported in terms of statistical significance as well as business language (practical significance), which business experts can understand and interpret. In this thesis, statistical results were expressed as effect sizes like Cohen‟s d that puts the results into standardised and measurable units, which can be reported practically. These effect sizes indicated strength of correlations between variables, contribution of variables to the odds of defaulting, the overall goodness-of-fit of the models and the models‟ discriminating ability between high and low risk customers. / Statistics / M. Sc. (Statistics)

When species compete for resources, in a stable homogeneous environment, there are two possible outcomes. The first is that one species will out-compete the other and exclude it from the environment. This is known as the competitive exclusion principle. The second is that both species will manage to coexist. Coexistence can only occur if the species’ niches are differentiated such that interspecific competition is minimised, or eliminated. This outcome is known as resource partitioning. Two closely related Australian skink species of the Egernia genus, Egernia multiscutata and Egernia whitii, are abundant and sympatric on Wedge Island in South Australia’s Spencer Gulf. The species are morphologically very similar and appear to have very similar life histories and habitat requirements. Ostensibly, they would compete for limiting resources in this environment. This thesis is the first investigation into resource partitioning in this previously unstudied model organism. I report the results of multi-faceted investigations into the coexistence of the skinks, E. multiscutata and E. whitii on Wedge Island and the evidence for, and mechanisms of, any facultative resource partitioning between them. Study methods involved a transect survey of most of Wedge Island to determine the species’ distributions and any evidence for resource partitioning; a morphological comparison to investigate any potential competitive advantages of either species; a habitat choice experiment to establish retreat-site preferences in the absence of interspecific interference; and, a series of staged dyadic encounter experiments to investigate interspecific competitive interactions. Resource partitioning was evidenced by differential distributions of the species among substrates containing the elements required for permanent refuge shelters. This partitioning was not mediated by avoidance of particular substrates but by the presence of the opponent species, combined with attraction to suitable substrates. Asymmetries in some morphological characters were found to confer a potential competitive advantage to E. multiscutata in agonistic encounters with E. whitii. Both species were found to have the same refuge site preferences when interference competition was experimentally removed. This result was not concordant with observed resource partitioning in the field and suggests that the habitat choices of both species are modified by the presence of the opponent species. Analyses of staged dyadic encounter experiments showed that E. multiscutata was more likely to gain greater access to a contested habitat resource and more likely to exclude E. whitii from the resource than vice-versa. Nevertheless, the outcome of competitive interactions was not completely deterministic and there was some tolerance of co-habitation. E. multiscutata’s competitive advantage was attributable largely to its greater mass and head dimensions relative to snout to vent length. However, differential behavioural responses to the threat of larger opponent size also played an important part in resource partitioning between the species.

企業發生財務危機，不僅使其經營陷入生死關頭之掙扎，更影響眾多投資人、債權人的利益，對於整個經濟環境亦造成一定的衝擊。因此，如何提早察覺企業之危機，以減少社會成本，實值得我們深入研究。 本研究主要目的為評估現金流量表揭露之資訊，於預測企業財務危機的有用性。本研究欲探討現金流量資訊是否為預測企業財務危機的良好指標，於建構企業財務危機預警模式之際，加入現金流量的財務指標是否會比僅以傳統財務比率建立之預警模式，更具預測能力。 本研究採用配對樣本設計，在我國上市公司中共選取了35家危機公司與68家正常公司。並利用Logit迴歸分析分別建立現金流量模式、應計財務模式與綜合模式，得到以下結論： 一、在財務危機發生之前一至三年，本研究所使用的應計基礎財務比率並非皆適合用來區分危機公司與正常公司。 二、除了營業活動現金流量相關比率具有顯著的區別能力外，部分投資與融資活動現金流量相關比率亦提供額外的財務危機警訊。 三、現金流量比率預警模式之預測力表現不遜於應計基礎比率模式；但在應計基礎比率中加入現金流量比率，並未顯著提高模式的預測能力。 / The objective of this study is to assess the usefulness of cash flow disclosures in the prediction of financial distress. This study also determines whether cash flow ratios are good indicator of financial distress and whether adding cash flow ratios in prediction model can improve the predictive ability of the model employing conventional accrual-based ratios. Using a matched pair design, this study examines a sample of 35 distress firms along with 68 non-distress firms. Also, a logistic regression analysis is used to establish the financial distress model with and without cash flow variables respectively, in order to test the hypotheses developed by this study and to derive the conclusion. The findings of this study are as follows. 1. During the period between 3 years to 1 year before financial distress, the accrual-based ratios used in this study aren't all good predictor in financial distress model. 2. The discriminate ability of operating cash flow data is significant. Also, the investing and financing cash flow data provide additional information in the prediction of business distress. 3. Cash flow ratios provide a superior measure for the prediction of financial distress over accrual-based ratios. However, no significant evidence shows that using cash flow ratios in conjunction with accrual-based ratios can improve the overall predictive power of accrual-based ratios alone.

多樣化壽險商品時代來來臨、壽險業目標市場鎖定的需求，致使「選擇適當的目標市場」和「設計適合目標市場的行銷組合策略」成為保險業者在擬定行銷策略所應注重兩大方向。惟如何選擇最具有吸引力，而又適合本身資源條件及競爭環境的目標市場，則就是行銷理論研究及實務上最重要的一項問題。 綜觀目前國內各研究所的論文中，關於消費者對於保險商品之購買行為的研究，多集中在消費者購買保險的原因或動機之分析上。惟其多是描述性、相關性分析為主，而此類研究方式雖然有其實用性，但其在缺乏「因果關係」的分析下，實無法了解消費者本身之不同，所引起購買意願之不同。再者，其並未進一步針對不同商品，研究影響消費者之所以購買不同商品之因素，係因任何忽略「商品多樣性」的研究，顯然過於簡化影響是否消費者購買保險商品之因素。本文對於消費者購買行為之基本認識為：消費者決定『是否購買』。保險，以及決定『選擇何種』保險的過程是同一的、不可分的。」因此，本文將以「多樣化的保險商品」為前提，來研究消費者決定「是否」購買保險、以及「選擇」購買何種保險之原因或動機。 究竟「消費者本身的差異性」與「是否購買及購買何種保險」之間存在什麼樣的關係，同時也是保險公司在保單設計、搭配、以及保險行銷上不可忽略之重要裁題。本文從于證資料上分析此一問題，以EKB消費者行為模式為理論基礎，依消費者本身的不同的背景、不同投保的動機、不同對保險的認知等等不同追求產品的相對利益為基礎因素，來探討消費者對不同的保險商品的需求。 本文乃以『問卷調查』為研究工具，針對台灣地區 20 歲- 70 歲之消費者為研究對象，實際訪查消費者所偏好之保險商品。共計取得有效問卷 965 份，輔以以效用函數為理論基礎之『羅吉特迴歸模式』計量方式，找出「消費者本身的差異性」與「是否購買及購買何種保險產品」之間的因素，並而建立消費者效用函數，進而預估消費者購買保險機率，促使業者更能設計符合消費者需求之保單組合。 研究結果顯示，在「是否會購買保險」的議題上，發現消費者教育程度不同會影響其購買意願；「保險演講會的舉行」、「親友在保險公司做事」或「自身或家屬曾發生事故」時，亦會明顯提高消費者之購買意願；業務員的上門推銷將是促使消費者引起購買保險的主要動機之一；再者，消資者在購買保險時，最重要之評估準則，則在於壽險期間是否太長、領回的錢值不值得及保費是否會太高等問題。 就「偏好購買不同商品」的議題上，本研究亦就目前市面上較為普遍之十種商品作研究，研究發現影響消費者偏好購買各個商品之因素各有不同，本文亦對其作綜合整理。最後亦針對研究結果，就各個不同保險商品，依其具有顯著水準之人口統計變項作--市場區隔，以期能提供保險公司或業務員在銷售時，可依商品的不同對消費大眾做市場區隔，使業務員或保險公司較易針對消費者不同的需求，做出較適合消費者且較易使消費者接受的保單設計。相信如此一來，非但有助於保險公司之保單設計與行銷，對於保險消費者如何選擇最適合自己的保單，也有相當的助益。 / The main goal of this research is to study the motives of the consumers purchasing the insurance policies and the selecting procedures. The previous researches on this area have been focused on the purchasing motives of each individual consumer. This kind of approach is widely used in practice. However, the consumers are not facing one insurance product but a variety of different insurance portfolios. In this study, we focus on analyzing the consumer-purchasing behavior of insurance portfolios. The logistic regression model is used to estimate the preference of the consumers among different insurance policies. The procedures of this study are summarized in the following: ( 1 )Review the developments of the previous researches and the findings. ( 2 )Design an appropriate questionnaire to collect the valid information and formulate the logistic regression model in this study. ( 3 )Collect the samples from the questionnaire and code this survey into a database system. ( 4 )Estimate the coefficients in the regression model in Step (2) and analyze the results. Finally comment on the findings. Using the logistic regression model is helpful for the marketing department in insurance company to target the appropriate populations and differentiate the various insurance portfolios. In this study, the information from the questionnaire is investigated based on our choice model. Monitoring these effects is beneficial for the managers having concise information in our target markets. Finally, a quantitative model is proposed for Taiwan insurance markets and the recommended marketing strategy.