This dissertation addresses problems found in observational cohort studies where the repeated outcomes of interest are truncated by both death and by dropout. In particular, we consider methods that make inference for the population of survivors at each time point, otherwise known as 'partly conditional inference'. Partly conditional inference distinguishes between the reasons for missingness; failure to make this distinction will cause inference to be based not only on pre-death outcomes which exist but also on post-death outcomes which fundamentally do not exist. Such inference is called 'immortal cohort inference'. Investigations of health and cognitive outcomes in two studies - the 'Origins of Variance in the Old Old' and the 'Health and Retirement Study' - are conducted. Analysis of these studies is complicated by outcomes of interest being missing because of death and dropout. We show, first, that linear mixed models and joint models (that model both the outcome and survival processes) produce immortal cohort inference. This makes the parameters in the longitudinal (sub-)model difficult to interpret. Second, a thorough comparison of well-known methods used to handle missing outcomes - inverse probability weighting, multiple imputation and linear increments - is made, focusing particularly on the setting where outcomes are missing due to both dropout and death. We show that when the dropout models are correctly specified for inverse probability weighting, and the imputation models are correctly specified for multiple imputation or linear increments, then the assumptions of multiple imputation and linear increments are the same as those of inverse probability weighting only if the time of death is included in the dropout and imputation models. Otherwise they may not be. Simulation studies show that each of these methods gives negligibly biased estimates of the partly conditional mean when its assumptions are met, but potentially biased estimates if its assumptions are not met. In addition, we develop new augmented inverse probability weighted estimating equations for making partly conditional inference, which offer double protection against model misspecification. That is, as long as one of the dropout and imputation models is correctly specified, the partly conditional inference is valid. Third, we describe methods that can be used to make partly conditional inference for non-ignorable missing data. Both monotone and non-monotone missing data are considered. We propose three methods that use a tilt function to relate the distribution of an outcome at visit j among those who were last observed at some time before j to those who were observed at visit j. Sensitivity analyses to departures from ignorable missingness assumptions are conducted on simulations and on real datasets. The three methods are: i) an inverse probability weighted method that up-weights observed subjects to represent subjects who are still alive but are not observed; ii) an imputation method that replaces missing outcomes of subjects who are alive with their conditional mean outcomes given past observed data; and iii) a new augmented inverse probability method that combines the previous two methods and is doubly-robust against model misspecification.

This thesis is a study of the coerciveness of legal systems. I defend two main claims: that typical legal systems are much less coercive than most legal and political philosophers think, and that legal systems are not necessarily coercive. My defence is developed in three parts. The first is dedicated to building the necessary theoretical framework to defend the main claims of this thesis. This is where I offer a rigorous formulation of the questions that this thesis addresses and contextualise them within broader debates about the relationship between law and coercion. A substantial portion of the first part of the thesis is devoted to the development of two accounts: an account of coercion and an account of the conditions legal systems must satisfy in order to be coercive. The second part is where I advance two arguments for the claim that typical legal systems are much less coercive than it is usually thought. The first is an argument that establishes that our legal systems rarely issue conditional threats. Given that issuing conditional threats is a necessary condition for any legal system to be coercive - or so I claim in the first part of the thesis - the fact that our legal systems rarely do so undermines the view that our legal systems are pervasively coercive. The second argument is based on the reasons why citizens comply with legal mandates. I analyse the relevant empirical data and show that compliance is not frequently owed to the threat of unwelcome consequences. This should not have been the case had our legal systems been as coercive as philosophers generally think. The third part deals with the claim that legal systems are necessarily coercive. There I address some methodological concerns that this claim gives rise to and propose two arguments for viewing coerciveness as a contingent feature of our legal systems.

Objetivo: Apresentar e discutir métodos para estimar prevalências ajustadas em pesquisas clínicas e epidemiológicas, bem como desenvolver rotinas computacionais em SAS e R. Métodos: No contexto de estudo transversal, foi simulada uma amostra de 2.000 observações independentes, considerando o desfecho dicotômico diabetes, sexo como a variável de exposição e idade como variável de ajuste. As estimativas de prevalências ajustadas (IC 95%) foram estimadas pelos métodos de predição condicional e marginal, utilizando as rotinas desenvolvidas em SAS e R. O método Delta foi usado para construir os intervalos de confiança. Os resultados foram comparados com aqueles do SUDAAN (SAS-Callable), Stata e a macro %ADJ_PROP (SAS). Resultados: No exemplo simulado, 68,2% são do sexo feminino e a idade média (DP) foi 57,6 (5,0) anos, sendo 54,2 (3,9) anos em homens e 59,2 (4,6) anos em mulheres. A estimativa da prevalência global do desfecho foi de 25,3% (IC 95%:23,4-27,3); sendo 13,8% (IC 95%:11,7-16,7) e 30,7% (IC 95%:28,3-33,2), respectivamente para homens e mulheres. As estimativas de prevalências ajustadas por idade, por meio do método de predição condicional, foram de 19,6% (IC 95%:16,2-23,6) para homens, e 23,6% (IC 95%:21,2-26,1) para mulheres. Pelo método de predição marginal, as estimativas foram de 22,4% (IC 95%:18,7-26,5) para homens, e 26,3% (IC 95%:24,1-28,6) para mulheres. Conclusão: A discrepância entre as estimativas não ajustadas é devida ao confundimento pela idade. Estimativas livres de confundimento podem ser obtidas por meio das prevalências ajustadas pela idade. No entanto, a estimativa pelo método de predição condicional não engloba a prevalência global. Em virtude disso, o método de predição marginal é, geralmente, mais adequado. A rotina desenvolvida na versão para R é uma alternativa aos softwares comerciais. / Objective: To present and discuss methods to estimate adjusted prevalences for clinical and epidemiological research, and develop computational routines in SAS and R. Methods: In the context of cross-sectional study, it was simulated a sample of 2,000 independent observations, considering the dichotomous outcome diabetes, sex as the exposure variable and age as an adjustment variable. Adjusted prevalences were estimated by the conditional and marginal methods, using routines developed in SAS and R. Confidence intervals were constructed using the Delta method. The results were compared with those of the SUDAAN (SAS-callable), Stata and macro %ADJ_PROP (SAS). Results: In simulated example, 68.2% are female and the mean (SD) age was 57.6 (5.00) years old, being that 54.2 (3.94) years for men and 59.2 (4.60) years in women. The estimated global prevalence of outcome was 25.3% (CI 95%: 23.4-27.3) and 13.8% (CI 95%: 11.7-16.7) and 30.7% (CI 95%: 28.3-33.2), respectively for men and women. Estimates of adjusted prevalence for age, through the conditional method, were 19.6% (CI 95%: 16.2-23.6) for men, and 23.6% (CI 95%: 21,2-26.1) for women. For marginal method, the estimates were 22.4% (CI 95%: 18.7-26.5) for men and 26.3% (CI 95%: 24.1-28.6) for women. Conclusion: The observed discrepancy in estimates by sex, unadjusted, can be attributed to confounding due to difference in age distribution between sexes. Comparable estimates (without confounding) of the prevalences can be obtained through prevalence adjusted for age. However, the estimate for the conditional method does not comprise the global prevalence. As a result, the marginal method is in general more suitable. The developed routines can be useful for estimating adjusted prevalences, particularly the R version (an alternative to commercial software).

Orientador: Maria de Lourdes Morales Horiguela / Banca: Tânia de Rose / Banca: Tania Moron Saes Braga / Banca: Kester Carrara / Banca: Maria Regina Cavalcante / Resumo: Os objetivos deste trabalho foram os de capacitar professores da sala de reforço, do Ensino Fundamental, para a utilização da metodologia da equivalência de estímulos com seus alunos, por meio de um programa informatizado e, por meio da aplicação da metodologia da equivalência, verificar a história de interação de dois alunos com as figuras referentes às palavras de treino e comparar o desempenho deles com os demais, quanto à leitura generalizada. Participaram três professoras do Ensino Fundamental (sendo duas da mesma escola) e quatro alunos, dois de cada escola, apresentando dificuldades de leitura e escrita, como omissão da leitura e/ou da escrita de sílabas, troca de sílabas ou não conseguir ler e/ou escrever nem parte da palavra. As escolas ficam situadas em Londrina, Paraná. Inicialmente realizou-se a seleção dos participantes, que ocorreu por meio de indicação das supervisoras, um pré-teste de leitura com os alunos para estabelecer a linha de base e o pré-treino. No pré-treino, foram apresentadas figuras correspondentes a cada palavra que seria ensinada (palavra de treino) e solicitado à criança que dissesse o nome da figura e, para dois alunos, um de cada escola, que leram menos palavras no pré-teste, foi pedido que falassem o que conheciam sobre cada figura que estava sendo apresentada. Paralelamente, as professoras participaram de encontros para a discussão de questões teóricas relativas à metodologia, bem como para conhecer o software e aprender a manuseá-lo. Em seguida, conduziram as sessões experimentais com os alunos para a aplicação da metodologia de equivalência de estímulos, para o ensino de leitura de palavras. / Abstract: The objectives of this work were the ones of qualifying teachers of the reinforcement room, of the Fundamental Teaching, for the use of the methodology of the equivalence of incentives with your students, through a computerized program and, through the application of the methodology of the equivalence, to verify the history of two students' interaction with the referring illustrations to the training words and to compare their acting with the others, with relationship to the widespread reading. Announced three teachers of the Fundamental Teaching (being two of the same school) and four students, two of each school, presenting reading and writing difficulties, as omission of the reading and/or of the writing of syllables, change of syllables or not to get to read and/or to write nor it a part of the word. The schools are located in Londrina, Paraná. Initially was realized the participants' selection, that it happened through the supervisors' indication, a reading initial test with the students to establish the base line and the initial training. In the initial training, were presented illustrations corresponding to each word that would be taught (training word) and requested the child that said the name of the illustration and, for two students, one of each school, that read less words in the initial read, was asked for they spoke what knew about each illustration that was being presented. Parallel to that, the teachers participated of encounters for the discussion of relative theoretical subjects to the methodology, as well as to know the software and to learn to handle. Soon after, they led the experimental sessions with the students for the application of the methodology of equivalence of incentives, for the teaching of reading of words. / Doutor

Os empreendimentos de mineração comumente demandam grande quantidade de investimentos financeiros e, na maioria das vezes, longos períodos de implantação, o que os torna altamente sujeitos a diversas fontes de incertezas. Tais incertezas comumente tendem a diminuir conforme a evolução do projeto. O objetivo deste estudo é correlacionar as incertezas associadas ao modelo de teores de cobre do depósito Sequeirinho com o volume de investimentos realizados ao longo de distintas fases da pesquisa geológica. Este depósito insere-se no contexto do Complexo de Mineração Sossego, localizado no município de Canaã dos Carajás (PA). Primeiramente, foram realizadas 100 simulações para cada domínio litológico em cada campanha de sondagem (pré-1998, 1999, 2000, 2002 e 2003) a partir do método de simulação sequencial gaussiana condicionada aos dados amostrais, totalizando 1.400 possíveis cenários. Para a avaliação das incertezas foram calculados três índices: variância condicional, coeficiente de variação condicional e intervalo interquartil. Por fim, a avaliação dos investimentos foi elaborada a partir dos custos estimados para o desenvolvimento de sondagens e análises químicas. Desde a campanha pré-1998, houve uma tendência de os teores médios do depósito aproximarem-se dos prováveis valores reais observados nas fases finais da pesquisa. No ano de 2000 ocorreu o maior investimento (cerca de 28 milhões de Reais) e a redução das incertezas atingiu o patamar de 15%. Os investimentos desenvolvidos em sondagens posteriores à campanha de 2000 foram da ordem de 9 milhões de Reais (cerca de 12 mil metros de sondagem), porém, não foram constatadas reduções significativas das incertezas. Este investimento seria melhor aproveitado caso fosse redirecionado a novas áreas de prospecção. Além do montante financeiro necessário para a redução das incertezas, foco deste estudo, as variações na interpretação geológica e a locação dos furos de sondagem são variáveis importantes na análise de incertezas associadas aos investimentos em pesquisa geológica. / Mining projects require large amount of capital investment and most cases spend long periods of implementation, which make them extremely susceptible to several sources of uncertainty. Such uncertainties commonly tend to decrease along the project development. The present study aims to correlate the uncertainties associated to the grade model of the Sequeirinho copper mine with the amount of capital investment that has been spent along different geological surveys. Sequeirinho site is included in the context of Sossego Mine Complex, located in Canaã dos Carajás (PA, Brazil). Initially, 100 simulations were carried out for each lithologic domain in each drilling campaign (pre-1998, 1999, 2000, 2002 and 2003) using sequential Gaussian simulation conditioned to the sample, resulting in 1,400 possible scenarios. Three indexes were calculated for the uncertainty analysis: conditional variance, conditional coefficient variance and interquartile range. Finally, the evaluation of capital investment was elaborated from the costs estimated for drilling and chemical analysis. Since pre-1998 drilling campaign, deposit average grades have showed a trend to get closer to the possible real values observed in the final research surveys. In 2000, there was the biggest capital investment (about R$ 28 millions) and the uncertainty reduction has reached the maximum level of 15%. Investments performed in drilling programs after 2000 were around R$ 9 million (around 12,000 meters of drill holes), but the uncertainty reduction was not significant. Therefore, such investments might be used to discover new targets. Besides the correlation of uncertainty reduction and amount of capital investment, the main focus of this study, the uncertainty of geological model and the drillhole locations are important variables to be considered in investment analysis of geological survey.

Submitted by Erika Demachki (erikademachki@gmail.com) on 2014-09-02T16:41:54Z No. of bitstreams: 2 TCC_Vanessa - 5 (5).pdf: 2109952 bytes, checksum: 2b62e1742f30b835be29c5f2cd7ec5ae (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) / Made available in DSpace on 2014-09-02T16:41:54Z (GMT). No. of bitstreams: 2 TCC_Vanessa - 5 (5).pdf: 2109952 bytes, checksum: 2b62e1742f30b835be29c5f2cd7ec5ae (MD5) license_rdf: 23148 bytes, checksum: 9da0b6dfac957114c6a7714714b86306 (MD5) Previous issue date: 2013-03-01 / This work aims to establish links between the di erent possible ways to resolve some issues involving the calculation of the probability of an event occurring through the use of the Multiplication of Probability Theorem (Theorem 2) and, from it, to show that we can solve problems involving simultaneous withdrawals, originally settled by Combinatorial Analysis, replacing them by successive withdrawals without replacement, considering the range of possible clusters (Theorem 4). Highlighting some precursors and their contributions to the development of Probability Theory, such as Cardano, Pascal, Laplace and Kolmogorov, among others. Exemplify, through problem solving, application of Theorem 4 to show the simpli cation of the calculation of probabilities applied, and propose activities that encourage discussion in the classroom, in order to clarify with the teacher's intervention, concepts such as Independent Events and using the Binomial Distribution. / Este trabalho tem como objetivo estabelecer as ligações entre as diferentes formas possíveis de resolver algumas questões envolvendo o cálculo da probabilidade de um evento ocorrer, mediante o uso do Teorema da Multiplicação das Probabilidades (Teorema 2) e, a partir dele, demonstrar que podemos resolver problemas que envolvam retiradas simultâneas, inicialmente resolvidos por Análise Combinatória, substituindo-as por retiradas sucessivas e sem reposição, considerando a ordem dos grupamentos possíveis (Teorema 4). Destacar alguns precursores e suas contribuições para o desenvolvimento da Teoria das Probabilidades, tais como, Cardano, Pascal, Laplace e Kolmogorov, dentre outros. Exemplifi car, através da resolução de problemas, aplicações do Teorema 4 para mostrar a simpli cação dos cálculos das probabilidades pedidas, bem como propor atividades que favoreçam debates em sala de aula, com o objetivo de clarifi car, com a intervenção do professor, conceitos como Eventos Independentes e o uso da Distribuição Binomial.

As classes de equivalência de estímulos são produto das contingências de reforço. No entanto, as pesquisas têm utilizado apenas contingências de reforço positivo e investigado, na sua maioria, como eventos relacionados a um mesmo reforçador podem passar a compor uma mesma classe. O objetivo do presente trabalho foi avaliar o efeito de diferentes contingências de reforço, positivo e negativo, no estabelecimento de discriminações condicionais na formação de classes de equivalência em humanos. Foram realizados dois experimentos. No Experimento I, 12 participantes treinaram simultaneamente 12 relações condicionais (AB e BC) utilizando três arranjos de contingência de reforçamento diferentes: reforçamento positivo (Ganhar/Manter), reforçamento negativo (Manter/Perder) e uma contingência mista de reforçamento negativo e positivo (Ganhar/Perder), seguidos de testes de formação de classes de equivalência. No Experimento I, os participantes expostos aos testes de equivalência acertaram todas as tentativas de teste. Estes resultados demonstram a possibilidade de formação de classes de equivalência em uma contingência de reforço negativo, porém um possível efeito de teto impede a comparação de cada contingência na formação de classes. No Experimento II, sete participantes foram expostos ao treino de 12 relações condicionais, com classes com um maior número de nódulos (AB, BC e CD) utilizando apenas duas contingências de reforço (Ganhar/Manter e Manter/Perder). Em ambos os experimentos, a sequência de aquisição das discriminações e o viés inicial produzido pelas contingências de reforço negativo sugerem um impacto diferencial da punição em relação ao reforçamento. Os resultados nos testes de equivalência sugerem que a formação de classes é menos provável quando utilizada uma contingência de reforço negativo em relação a de reforço positivo. Foi discutida a generalidade da formulação sobre formação de classes de equivalência, priorizando os padrões de controle de estímulos produzido pelas diferentes contingências. Ressalta-se a necessidade de maior investigação sobre quais os efeitos do uso de contingências aversivas no controle da resposta e mais especificamente no controle de estímulos / Equivalence classes are produced by reinforcement contingencies, however, research have used exclusively positive reinforcement and investigated mostly how events related to a same reinforcer may result in the formation a same class. The aim of this study was to evaluate how different reinforcement contingencies affect the acquisition of conditional discriminations and equivalence class formation in humans. Two experiments were conducted. In Experiment I, 12 participants learned 12 conditional relations (AB BC) under three different contingencies: positive reinforcement (Gain/Maintain), negative reinforcement (Maintain/Lose) and a mixed contingency (Gain/Lose). Equivalence class test were applied subsequently. In Experiment I, participants that underwent the equivalence tests performed correctly in all trials, although a roof effect might have prevented a proper comparison of the contingencies effect. In Experiment II, seven participants learned 12 conditional relations with more nodes (AB, BC e CD) and two reinforcement contingencies (Gain/Maintain and Maintain/Lose). In both experiments, the conditional discrimination acquisition sequence and the initial bias produced by the negative reinforcement contingency suggest a differential effect of punishment in relation to reinforcement. The equivalence test results suggest the possibility of equivalence class formation using a negative reinforcement contingency and indicate that equivalence classes are less probable using a negative reinforcement contingency than a positive reinforcement one. The generality of the equivalence class formulation was discussed emphasizing the investigation of stimuli control patterns produced by the contingencies. The lack of knowledge about stimulus control produced by aversive contingencies was highlighted

Probabilistic models of text are a useful tool for enabling the analysis of large collections of digital text. For example, Latent Dirichlet Allocation can quickly produce topical summaries of large collections of text documents. Many important uses cases of such models include human interaction during the inference process for these models of text. For example, the Interactive Topic Model extends Latent Dirichlet Allocation to incorporate human expertiese during inference in order to produce topics which are better suited to individual user needs. However, interactive use cases of probabalistic models of text introduce new constraints on inference - the inference procedure must not only be accurate, but also fast enough to facilitate human interaction. If the inference is too slow, then the human interaction will be harmed, and the interactive aspect of the probalistic model will be less useful. Unfortunately, the most popular inference algorithms in use today either require strong approximations which can degrade the quality of some models, or require time-consuming sampling. We explore the use of Iterated Conditional Modes, an algorithm which is able to obtain locally optimal maximum a posteriori estimates, as an alternative to popular inference algorithms such as Gibbs sampling or mean field variational inference. Iterated Conditional Modes algorithm is not only fast enough to facilitate human interaction, but can produce better maximum a posteriori estimates than sampling. We demonstrate the superior performance of Iterated Conditional Modes on a wide variety of models. First we use a DP Mixture of Multinomials model applied to the problem of web search result cluster, and show that not only can we outperform previous methods in clustering quality, but we can achieve interactive runtimes when performing inference with Iterated Conditional Modes. We then apply Iterated Conditional Modes to the Interactive Topic Model. Not only is Iterated Conditional Modes much faster than the previous published Gibbs sampler, but we are better able to incorporate human feedback during inference, as measured by accuracy on a classification task using the resultant topic model. Finally, we utilize Iterated Conditional Modes with MomResp, a model used to aggregate multiple noisy crowdsourced data. Compared with Gibbs sampling, Iterated Conditional Modes is better able to recover ground truth labels from simulated noisy annotations, and runs orders of magnitude faster.

Psychological biases in the form of sentiment can affect various economic decisions including accounting choices. Broadly defined, the term sentiment refers to unjustified beliefs about the future cash flow prospects of the firm (Baker and Wurgler 2006). Asymmetric timely loss recognition (ATLR) is particularly prone to managerial sentiment because the decision to recognize economic gains and losses is based, in part, on managers’ beliefs about the likelihood of future economic events affecting the firms. In this study, I examine the effect of psychological biases about future performance on current accounting choices via the effect of market-level managerial sentiment on ATLR. I find that ATLR decreases with managerial sentiment and that periods of high managerial sentiment are associated with lower concurrent write-offs but higher subsequent write-offs. This study enhances the implications of sentiment on firms’ accounting choices by identifying a time-varying macroeconomic determinant of ATLR that is based on psychological biases about future performance.

To ensure test security and fairness, alternative forms of the same test are administered in practice. However, alternative forms of the same test generally do not have the same test difficulty level, even though alternative test forms are designed to be as parallel as possible. Equating adjusts for differences in difficulties among forms of the test. Six traditional equating methods are considered in this study: equipercentile equating without smoothing, equipercentile equating with pre-smoothing and post-smoothing, IRT true-score and observed-score equatings, and kernel equating. A common feature of all of the traditional procedures is that the end result of equating is a single transformation (or conversion table) that is used for all examinees who take the same test. Van der Linden has proposed conditional equipercentile (or local) equating (CEE) to reduce the error of equating contained in the traditional equating procedures by introducing individual level equating. Van der Linden's CEE is conceptually closest to IRT-T in that CEE is with respect to a type of true score (θ, or proficiency), but it shares similarities with to IRT-O in that CEE uses an estimated observed score distribution for each individual θ to equate scores using equipercentile equating. No real-data study has yet compared van der Linden's CEE with each of the traditional equating procedures. Indeed, even for the traditional procedures, no study has compared all six of them simultaneously. In addition to van der Linden's CEE, two additional variations of CEE are considered: CEE using maximum likelihood (CEE-MLE) and CEE using the true characteristic curve (CEE-TCC). The focus of this study is on comparing results from CEE vis-à-vis the traditional procedures, as opposed to answering a “best-procedure”question, which would require a common conception of “true”equating. Although the results of the traditional equating methods are quite similar, the kernel equating method and equipercentile equating with log-linear presmoothing generally show better fit to the respective original form statistical moments under various data conditions. Although IRT-T and IRT-O usually are found to be least favorable under all circumstance in terms of statistical moments, the equated raw score difference distribution illustrates more stable performance than traditional equating methods. It was found here that the number of examinees having a particular score point does not influence results for CEE as much as it does for traditional equatings. CEE-EAP and CEE-MLE are very similar to one another and the equated score difference distributions are similar to those of IRT-O. CEE-TCC involves a part of the IRT-T procedure. Hence, CEE-TCC behaves somewhat similar to IRT-T. Although CEE results are less desirable in terms of maintaining statistical moments, the equated score differences are more consistent and stable than for the traditional equating methods.