Submitted by Caroline Xavier (caroline.xavier@pucrs.br) on 2017-08-07T13:36:00Z No. of bitstreams: 1 DIS_OTAVIO_CANOZZI_CONCEICAO_COMPLETO.pdf: 1480241 bytes, checksum: 6c742bea6a2e5e1360c0f6751123d488 (MD5) / Made available in DSpace on 2017-08-07T13:36:00Z (GMT). No. of bitstreams: 1 DIS_OTAVIO_CANOZZI_CONCEICAO_COMPLETO.pdf: 1480241 bytes, checksum: 6c742bea6a2e5e1360c0f6751123d488 (MD5) Previous issue date: 2017-03-23 / Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior - CAPES / The present dissertation evaluates the Science without Borders program focusing on three outcome variables. To achieve this goal, the work is divided into two essays. The first essay addresses the causal impact of the program on the performance of CsF alumni from the 2013 National Student Performance Exam (ENADE) microdata. The results are compared with two control groups: the first composed by students who also made international academic mobility, but through other initiatives and the second by the non-exchange students. The strategy of comparing the CsF?s alumni with the exchange students aims to mitigate the problems of self-selection, because we can expect more similar characteristics between these groups, both with regard to motivation to study abroad and observable characteristics. The methodology used was the Propensity Score Matching (PSM), using different pairing algorithms to check the validity of the findings. The results show that the alumni of the program present better grades than the exchange students, both in the general formation part and in the specific component of the ENADE exam. The second essay examines dropout and completion rates in the CsF students? course from the microdata of the Higher Education Census of 2014 and 2015. For this purpose, a panel of students was constructed, divided into the same groups of the first essay, which allowed to identify their trajectories in the course. The methodology used involved two techniques: a multinomial logit and the Markov transition matrices. The evidence found shows that CsF students evades more the course and conclude it at lower rates in the expected time, according to the serial grade of the course, than both comparison groups. The results of the multinomial logit model, controlling characteristics of individuals, institutions, courses and regions, suggest that the program has a ceteris paribus effect on the extension of student time in the university. / A presente disserta??o avalia o programa Ci?ncia Sem Fronteiras (CsF) com foco em tr?s vari?veis de resultado. Para atingir esse objetivo, o trabalho est? dividido em dois ensaios. O primeiro ensaio trata do impacto causal do programa sobre o desempenho dos ex-bolsistas do CsF a partir dos microdados do Exame Nacional de Desempenho dos Estudantes (ENADE) de 2013. Os resultados s?o analisados comparativamente a dois grupos de controle: o primeiro composto por estudantes que tamb?m realizaram interc?mbio acad?mico internacional, por?m por interm?dio de outras iniciativas e o segundo por n?o intercambistas. A estrat?gia de compar?-los com os intercambistas visa mitigar os problemas de autossele??o, porque se espera caracter?sticas mais semelhantes entre esses grupos, tanto no que tange ? motiva??o para estudar no exterior quanto em caracter?sticas observ?veis. A metodologia empregada foi o Propensity Score Matching (PSM), sendo utilizados diferentes algoritmos de pareamento para conferir validade aos achados. Os resultados revelam que os ex-bolsistas do programa apresentam melhores notas que os intercambistas, tanto na parte de forma??o geral quanto no componente espec?fico da prova do ENADE. O segundo ensaio analisa a evas?o e a taxa de conclus?o no curso de alunos do CsF a partir dos microdados do Censo da Educa??o Superior de 2014 e 2015. Para tanto, construiu-se um painel de estudantes, divididos de acordo com os mesmos grupos do primeiro ensaio, que permitiu identificar suas trajet?rias no curso. A metodologia utilizada envolveu duas t?cnicas: um logit multinomial e as matrizes de transi??o de Markov. As evid?ncias encontradas mostram que os alunos do CsF trancam mais o curso e se formam a menores taxas no tempo esperado, de acordo com a grade seriada do curso, do que ambos os grupos de compara??o. Os resultados do modelo logit multinomial, controlando caracter?sticas dos indiv?duos, das institui??es, cursos e regi?es, sugerem que o programa tem um efeito ceteris paribus sobre a prorroga??o do tempo do aluno na universidade.

Com o crescimento da população em áreas urbanas, o transporte coletivo assume um importante papel nos deslocamentos diários nas cidades brasileiras. No entanto, o grande volume de pessoas no entorno desses sistemas podem resultar em pontos críticos de segurança viária. Recentemente, as pesquisas sobre melhorias na segurança viária estão se direcionando para a investigação da relação entre o ambiente construído - isto é, padrões de uso do solo; desenho urbano; e sistemas de transporte - e acidentalidade. Porém, a relação entre os sistemas de transporte coletivo por ônibus e o ambiente construído ainda é pouco explorada. O ponto mais crítico em segurança identificado na literatura em sistemas de transporte coletivo são as estações de embarque e desembarque. Assim, esta dissertação tem como objetivo verificar a existência de relação entre elementos do ambiente construído com a ocorrência e severidade de acidentes viários no entorno de pontos de parada em sistemas prioritários para ônibus. Para isso, aplicou-se os modelos de regressão Binomial Negativo (NB), na análise de frequência de acidentes, e os modelos Logit Ordenado (OL) e Logit Multinomial (MNL), para a análise de severidade. No modelo de frequência de acidentes totais, identificou-se duas variáveis significativas e, no modelo estimado para atropelamentos, três variáveis. O modelo de severidade resultou em 19 variáveis significativas para acidentes totais e oito variáveis significativas para atropelamentos. As variáveis referentes ao envolvimento de motocicletas, automóveis, presença de interseção e uso diversificado do solo foram as mais recorrentes. O estudo possibilitou também avaliar as diferenças e vantagens entre os modelos ordenados e os não-ordenados. O modelo Logit Multinomial teve ajustes levemente melhores do que o Logit Ordenado. Ressalta-se, entretanto, que o ajuste do modelo não deve ser o único critério a se considerar na escolha de modelos para o estudo de severidade de acidentes. / Public transport has played an important role in daily commutes in Brazilian cities as population grows in urban areas. However, high volumes of pedestrians near those systems may result in critical safety issues. Recent studies on road safety improvements have sought for evidence on the relationship between built environment - land use patterns; urban design; and transportation systems - and traffic safety. Nevertheless, the relationship between traffic safety at bus priority systems and the built environment is still incipient. According to studies, the most critical safety issues related to public transport are the stations for boarding and alighting. Thus, this thesis aims to verify the relationship between built environment and incidence and severity of traffic crashes in the surroundings of bus priority systems. To do so, we applied Negative Binomial regression models (NB) for the frequency analysis, and the Ordered Logit (OL) and Multinomial Logit (MNL) models for severity analysis. NB models resulted in two significant variables for traffic crashes in general and three variables for pedestrian crashes. The severity model resulted in 19 significant variables overall crashes and eight significant variables for pedestrian crashes. Variables related to the involvement of motorcycles, automobiles, intersections and mixed land use were significant in majority of model estimations. The results allowed to evaluate the goodness of fit between ordered and unordered models. MNL had a slightly better adjustments compared to OL. Goodness of fit, however, should not be the only criterion for selecting a model to assess crash severity.

Após o fim da Segunda Guerra Mundial, teve início um processo de aumento da utilização dos automóveis nos grandes centros urbanos que modificou o desenho das cidades e o padrão de deslocamento nesses espaços. Uma das diversas consequências negativas da consolidação do automóvel como principal meio de transporte nas metrópoles foi o aumento do número de acidentes no trânsito, principalmente os que envolviam os usuários de transporte ativo – pedestres e ciclistas. A promoção dos modos de transporte não motorizados, como a bicicleta, é tida como ação complementar para reversão desse processo, porém, para tanto, é necessário melhorar a percepção de segurança sobre o modo cicloviário. A tomada de decisão por parte dos gestores do tráfego acerca das intervenções viárias que melhoram a segurança dos ciclistas carece de entendimento sobre os elementos que influenciam a ocorrência de acidentes envolvendo estes usuários. Desse modo, o trabalho desenvolvido nesta dissertação tem como objetivo a verificação da existência de relação entre os elementos urbanos construídos, os fatores socioeconômicos e outras possíveis fontes de risco com a frequência e severidade dos acidentes cicloviários na cidade de Porto Alegre O processo de investigação utilizou softwares de georreferenciamento para consolidar os dados coletados segundo as unidades de análise definidas. A frequência de acidentes foi modelada através de Modelo de Regressão Binomial Negativo e a severidade de acidentes foi modelada através de Modelo Logit Multinomial. O modelo de frequência de acidentes apresentou duas variáveis significativas. Em seguida, foi gerado um novo modelo de frequência de acidentes a partir da eliminação de possíveis fontes de discrepância dos dados. Os resultados indicaram que este segundo modelo estava melhor ajustado ao cenário estudado. O modelo de severidade de acidentes foi composto por 18 variáveis significativas. Por fim, os elementos que demonstraram maior prevalência sobre a acidentalidade de ciclistas foram a interação com automóveis, a presença de vias arteriais, a proximidade com áreas de interseção e o envolvimento de usuários vulneráveis (jovens e idosos em bicicletas) nos acidentes. / After the end of the World War II, many cities around the world have experienced an increase on car using which changed the way cities were planned and also mobility patterns. That process has generated some problems at urban areas like road accidents increasing, most of them related to active transport modes - pedestrians and cyclers. Non motorized transportation modes fomentation, for example cycling, emerges as an alternative to change this issue. However, before it happens, cyclers need to feel more safe on transit. Understanding the influence of the elements that affect cyclers's safety gives to decision makers and traffic managers a better possibility to make the right decisions on safety improvement. Thus, this thesis aims to verify the relationship between built environment, socialeconomic factors and other risk sources and frequency and severity of traffic crashes involving cyclists in Porto Alegre. By using GIS softwares, the collected data were consolidated according to the defined analysis units. Accident’s frequencies were modeled and the coefficients were estimated by using Negative Binomial Regression Model and the severity of accidents was modeled as a Multinomial Logit Model. The accident frequency model counted with two sginificant variables. Then, a new accident frequency model was estimated from the elimination of possible sources of data discrepancy. Results have shown that the second model fit better on the studied scenario. The accident severity model resulted in 18 significant variables. Finally, the elements that showed highest prevalences on cyclists safety were car interaction, presence of arterial roads, proximity to intersection areas and vulnerable cyclists users (young and old ones cycling) involvement on accidents.

In 1993, a travel study was conducted along the Wasatch front in Utah (Research Systems Group INC, 2013). The main purpose of this study was to assess travel behavior to understand the needs for future growth in Utah. Since then, the Research Service Group (RSG), conducted a new study in 2012 to understand current travel preferences in Utah. This survey, called the Residential Choice Stated Preference survey, asked respondents to make ten choice comparisons between two hypothetical homes. Each home in the choice comparison was described by different attributes, those attributes that were used are, type of neighborhood, distance from important destinations, distance from access to public transport, street design, parking availability, commute distance to work, and price. The survey was designed to determine the extent to which Utah residents prefer alternative household attributes in a choice selection. Each attribute contained multiple characteristic levels that were randomly combined to define each alternative home in each choice comparison. Those choices can be explained by Random Utility Theory. Multinomial logistic regression will be used to estimate changes in utility when alternative attribute levels are present in a choice comparison. Using the coefficient estimate for price, a marginal willingness to pay (MWTP) for each attribute level will be calculated. This paper will use two different approaches to obtain MWTP estimates. Method One will use housing and rent price to recode the price variable in dollar terms as defined in the discrete choice experiment. Method Two will recode the price variable as an average ten percent change in home value to extrapolate a one-time payment for homes. As a result, we found that it is possible to obtain willingness to pay estimates using both methods. The resulting interpretations in dollar terms became more relatable. Metropolitan planning organization can use these results to understand how residents perceive home value in dollar terms in the context of location-based attributes for homes.

Vehicle ownership and utilization have a profound influence on activity-travel patterns of individuals, vehicle emissions, fuel consumption, highway capacity, congestion and traffic safety. The influence could be further skewed by the diversity of the vehicle fleet. This thesis presents a detailed analysis of the 2001 National Household Travel Survey data to understand the vehicle ownership patterns, fleet mix, allocation and utilization in the context of household and person socio-demographic characteristics. Along with a rich descriptive analysis, models of vehicle ownership and utilization are estimated to distinguish four vehicle types; cars, SUVs (sport utility vehicles), vans and pickup trucks based on their ownership by households and utilization patterns by household members. The primary driver level vehicle utilization analysis provides insights into the extent of allocation of a vehicle to a single person. In addition to confirming many perceptions about the ownership, acquisition and utilization patterns of different types of vehicles, this analysis brings out some subtle differences and similarities among the vehicle types. The analysis results indicate a greater propensity to acquire and use larger vehicles such as minivans, sports utility vehicles and pickup trucks among certain socio-demographic segments of population. Increased ownership and use of vans and SUVs, and their usage as personal vehicles rather than just work vehicles warrants a need to revise vehicle type specific policies, transportation planning and control measures.

針對具有多項分配(multinomial distribution)母體的類別資料，貝氏分析通常採取Dirichlet分配作為其先驗分配(prior distribution)，但在很多實際應用時，卻會遭遇困難；例如，當我們欲推估各年齡層佔總勞動力人口之比例時，母體除具多項分配外，其相鄰類別之比例亦相對接近；換言之，此時母體為具有平滑性(smooth)的多項分配，若依然採用Dirichlet分配作為其先驗分配，則會因為Dirichlet分配本身不具有平滑的特性，因而在做貝氏分析時會產生困擾。對這個難題Dickey and Jiang於1998年提出一個解決之道，他們的理論是對Dirichlet分配作適當之調整，將經過線性轉換後之Dirichlet分配稱為過濾後Dirichlet分配(filtered-variate Dirichlet distribution)，以過濾後Dirichlet分配作為調整後之先驗分配。對於Dickey and Jiang提出的方法，我們重新以蒙地卡羅法(Monte Carlo method)求出貝氏解，同時也嘗試以類似Makov and Smith (1977)和Smith and Makov (1978)對混合分配(mixture distribution)所用之準貝氏方法(quasi-Bayesian method)來逼近貝氏解。而本文將由電腦模擬的方式，探討貝氏方法與準貝氏方法之執行結果，並且考察準貝氏方法之收斂行為，對準貝氏方法的使用時機提出建議。

Multinomial selection problem procedures are ranking and selection techniques that aim to select the best (most probable) alternative based upon a sequence of multinomial observations. The classical formulation of the procedure design problem is to find a decision rule for terminating sampling. The decision rule should minimize the expected number of observations taken while achieving a specified indifference zone requirement on the prior probability of making a correct selection when the alternative configurations are in a particular subset of the probability space called the preference zone. We study the constrained version of the design problem in which there is a given maximum number of allowed observations. Numerous procedures have been proposed over the past 50 years, all of them suboptimal. In this thesis, we find via linear programming the optimal selection procedure for any given probability configuration. The optimal procedure turns out to be necessarily randomized in many cases. We also find via mixed integer programming the optimal non-randomized procedure. We demonstrate the performance of the methodology on a number of examples. We then reformulate the mathematical programs to make them more efficient to implement, thereby significantly expanding the range of computationally feasible problems. We prove that there exists an optimal policy which has at most one randomized decision point and we develop a procedure for finding such a policy. We also extend our formulation to replicate existing procedures. Next, we show that there is very little difference between the relative performances of the optimal randomized and non-randomized procedures. Additionally, we compare existing procedures using the optimal procedure as a benchmark, and produce updated tables for a number of those procedures. Then, we develop a methodology that guarantees the optimal randomized and non-randomized procedures for a broad class of variable observation cost functions -- the first of its kind. We examine procedure performance under a variety of cost functions, demonstrating that incorrect assumptions regarding marginal observation costs may lead to increased total costs. Finally, we investigate and challenge key assumptions concerning the indifference zone parameter and the conditional probability of correct selection, revealing some interesting implications.

This paper develops a new approach to the analysis of longitudinal paired comparison data, where comparisons of the same objects by the same judges are made on more than one occasion. As an alternative to other recent approaches to such data, which are based on Kalman filter- ing, our approach treats the problem as one of multivariate multinomial data, allowing dependence terms between comparisons over time to be incorporated. The resulting model can be fitted as a Poisson log-linear model and has parallels with the quadratic binary exponential distribution of Cox. An example from the British Household Panel Survey illustrates the approach. (author´s abstract) / Series: Research Report Series / Department of Statistics and Mathematics

In this dissertation we developed theoretical models to analyse Gaussian and multinomial distributions. The analysis is focused on classification in high dimensional feature spaces and provides a basis for dealing with issues such as data sparsity and feature selection (for Gaussian and multinomial distributions, two frequently used models for high dimensional applications). A Naïve Bayesian philosophy is followed to deal with issues associated with the curse of dimensionality. The core treatment on Gaussian and multinomial models consists of finding analytical expressions for classification error performances. Exact analytical expressions were found for calculating error rates of binary class systems with Gaussian features of arbitrary dimensionality and using any type of quadratic decision boundary (except for degenerate paraboloidal boundaries). Similarly, computationally inexpensive (and approximate) analytical error rate expressions were derived for classifiers with multinomial models. Additional issues with regards to the curse of dimensionality that are specific to multinomial models (feature sparsity) were dealt with and tested on a text-based language identification problem for all eleven official languages of South Africa. / Thesis (M.Ing. (Computer Engineering))--North-West University, Potchefstroom Campus, 2009.

In this dissertation we developed theoretical models to analyse Gaussian and multinomial distributions. The analysis is focused on classification in high dimensional feature spaces and provides a basis for dealing with issues such as data sparsity and feature selection (for Gaussian and multinomial distributions, two frequently used models for high dimensional applications). A Naïve Bayesian philosophy is followed to deal with issues associated with the curse of dimensionality. The core treatment on Gaussian and multinomial models consists of finding analytical expressions for classification error performances. Exact analytical expressions were found for calculating error rates of binary class systems with Gaussian features of arbitrary dimensionality and using any type of quadratic decision boundary (except for degenerate paraboloidal boundaries). Similarly, computationally inexpensive (and approximate) analytical error rate expressions were derived for classifiers with multinomial models. Additional issues with regards to the curse of dimensionality that are specific to multinomial models (feature sparsity) were dealt with and tested on a text-based language identification problem for all eleven official languages of South Africa. / Thesis (M.Ing. (Computer Engineering))--North-West University, Potchefstroom Campus, 2009.