This mater thesis is focused on building predictive models. Their fundamental task is to provide an early-warning system, giving information about potential enterprise bankruptcy. The main essence and aim of the thesis is to create multivariate classification models by using discriminant analysis and logistic regression. Emphasis is put on their predictive accuracy, which is assessed for period of three years before bankruptcy declaration. Attempts to optimize classification thresholds in order to increase the initial accuracy are also made. Evaluating classification reliability of several existing models and performing profile analysis assessing predictive ability of univariate ratios were accomplished as well.

Geographic research on the Corn Belt and other regional landscapes of the central U.S. has not to date identified quantitatively the climatic, edaphic, topographic, and economic characteristics that determine rural land cover, and that therefore govern land cover change. Using the USDA/NASS Cropland Data Layer, this study identifies these characteristics using Multivariable Fractional Polynomials within a logistic regression framework. It maps the suitability distribution for corn, soybeans, spring and winter wheat, cotton, grassland, and forest land covers that dominate the central U.S., at a 56m resolution across 16 central U.S. states. The non-linear logistic regression models are successful in identifying determinants of land cover with relative operating characteristic (ROC) scores that range from 0.769 for soybeans to 0.888 for forest, with a combined corn/soybean model achieving an ROC of 0.871. For corn and soybean models, when prior land cover of a pixel is added, predictability and ROC scores increase substantially (0.07-0.10), indicating a strong temporal feedback in land cover dynamics. This process also aids in the delineation of fields from pixels. Adding neighboring land covers, however, improves predictability and ROC scores only slightly (0.014-0.019), indicating a weak spatial feedback mechanism. By including annual crop prices within the logit models, economically marginal cropland that comes into crop production only when prices are high is identified in a spatially-explicit manner. This capacity improves further analyses of economic and environmental impacts of policies that affect crop prices. The sustainability of current rural land use trends in the central U.S. is highly dependent on the ability to adapt to changing climatic conditions of the 21st century. As the climate begins to shift towards longer growing seasons, more erratic rainfall patterns, and overall warmer temperatures, there is potential for major impacts on seven major land covers of the central U.S. Suitability landscapes of individual land covers (corn, soybeans, spring and winter wheat, cotton, grasslands, and forests) were utilized to determine the influence of climate change on these landscapes. Twenty-seven climate change projection scenarios based on three global climate models, three representative concentration pathways, and three time periods were applied to the land cover suitability maps utilizing raster regression. The area now identified as the Corn Belt is projected to see a dramatic shift in the suitable climate with a potential for a 30 percent increase in summer growing degree days. While the area where conditions are suitable for corn, soybeans and spring wheat are all expected to decrease, winter wheat has the potential to increase in suitable area. In order to maintain current geographic patterns of crop production, corn would need to be adapted to higher temperatures.

The purpose of this research is to generate the most accurate model possible for predicting locations most suitable for new wind energy development using a case study of Iowa. With the United States poised for significant growth in electricity generation from wind and other renewable resources, this research can help developers, transmission planners, landowners, as well as academics in predicting optimal locations for development. Iowa currently has the second greatest installed wind energy capacity and highest percentage of energy generated from wind of all the states, and has excellent resources and central location for a high level of continued development. This research employs a variety of methods including traditional constraint mapping techniques, logistic regression, and a hybrid of both approaches in attempting to produce the best predictive model possible. The best performing logistic regression model with 11 variables correctly predicts 90.1% of overall observations. A variety of output maps are produced and analyzed, and many suggestions for future research in this area are presented.

This research was conducted at Tsakane, Kwa-Thema, and Duduza, Ekurhuleni Metropolitan Municipality of Gauteng Province in South Africa. The objective of the study was to determine the effectiveness of social protection as a way of alleviating poverty in the study areas. The study was expected to contribute to the body of knowledge in social protection services as a way of alleviating poverty in the study areas. The study attempted to contribute to answers to the following research questions: (i) What are the socio-economic characteristics of the recipients of social protection measures in the three areas of study? (ii) What are the perceptions of the respondents about the South African government’s social protection in their areas? (iii) What are the factors that influence the effectiveness of social protection? Stratified random sampling with a proportional representation method was employed to select 200 respondents. The data collection tool used was simple closed-ended questionnaires. Interviews were conducted face-to-face with respondents. Statistical Package for Social Sciences (SPSS) version 21 of 2012 was used to analyse the data. Both descriptive statistics and binary logistic regression were employed. The results of the analysis revealed that the significant variables that had an effect on social protection were: the location of the respondents; their gender; their level of education; the type of dwelling of the respondents; and their income outside farming. The study recommends that the significant variables that had an effect on social protection be considered when measures of social protection measures are implemented / College of Agriculture and Environmental Sciences / M. Sc. (Agriculture)

The theoretical part of this master thesis deals with creation and solution of financial distress and analysing classification models. In the practical part I defined own methods for financial distress prediction of company using discriminant analysis and logistic regression.

Um dos temas mais estudados na área de finanças, em especial na análise de créditos, são os modelos que buscam prever a capacidade das empresas em se manterem solventes. Via de regra, tais estudos buscam, dentre vários indicadores, aqueles que se mostram mais apropriados para realizar tal predição. Nesse trabalho propõe-se um outro olhar sobre os modelos de previsão. Partindo de modelos já consagrados na literatura, escolheram-se os indicadores contábeis mais utilizados, que foram comparados, através da Análise Discriminante e da Regressão Logística, com os indicadores oriundos do Modelo Dinâmico. O objetivo do estudo foi verificar se os indicadores do Modelo Dinâmico oferecem melhores resultados que os indicadores tradicionais. O trabalho se baseia numa amostra com 48 empresas, composta de 24 insolventes e as outras 24 ditas como saudáveis, tratadas como pares das insolventes, escolhidas dentro do mesmo setor econômico de cada uma das insolventes. Além disso, foi incluída no estudo a classificação de empresas de Fleuriet como variável qualitativa. Os resultados obtidos não apresentam evidências sobre a superioridade de um ou outro conjunto de indicadores, mas, os melhores resultados alcançados derivam da inclusão da classificação de empresas de Fleuriet, seja através da Análise Discriminante, seja através da Regressão Logística, conseguindo no melhor dos resultados, um percentual de acerto total de 83,3%. A análise minuciosa dos erros de classificação ensejou uma proposta de reordenação dos tipos de situação de liquidez originalmente propostos por Fleuriet. / One of the most studied topics in finance, particularly in credit analysis, are the models that seek to predict the ability of firms to remain solvent. Usually, such studies among various indicators, seek those who are most appropriate to perform such a prediction. In this work we propose a different view of the forecast models. Starting from models already established in the literature, were chosen the most used financial indicators, which were compared by discriminant analysis and logistic regression, with indicators derived from the Dynamic Model. The aim of the study was to determine whether the indicators of Dynamic Model offer better results than the traditional indicators. The work is based on a sample of 48 companies, consisting of 24 insolvent 24 and the other said to be healthy treated as pairs of insolvent chosen within the same economic sector of each insolvent. Furthermore, it was included in the study the classification of companies as Fleuriet qualitative variable. The results show no evidence of the superiority of one or another set of indicators, but the best results achieved derive from the inclusion of classification Fleuriet companies, either through discriminant analysis, logistic regression is through, achieving the best results a percentage of the total adjustment of 83.3 %. A thorough analysis of the classification errors gave rise a proposed reorganization of the types of liquidity situation originally proposed by Fleuriet.

Background: As high HIV transmission rates persist in Sub-Saharan Africa, the effect of wealth inequality rather than solely absolute wealth as a potential driver of the HIV epidemic has been given increased attention in recent research, but has not yet been investigated in the Nigerian setting. As, particularly in contexts of socioeconomic inequality, individuals may face barriers to both obtaining health-related knowledge and translating this knowledge into actual engagement in preventive measures, it is relevant to assess the level of HIV-related knowledge in the Nigerian population. Furthermore, it is of interest to investigate its socioeconomic predictors, and to identify risk-groups for low HIV-related knowledge, which consequentially are also potential risk groups for high HIV transmission. This will ultimately facilitate the targeting and implementation of more appropriate and effective preventive interventions among these groups. Due to the country’s high HIV prevalence and its ethnic and socioeconomic heterogeneity, it is both an interesting and highly relevant setting in which to analyse the socioeconomic determinants of HIV-related knowledge. Methods: Utilizing data from the Nigerian Demographic and Health Survey, Paper 1 of this thesis investigates wealth inequality as a predictor of low HIV-related knowledge in the Nigerian population through logistic regression modeling. The effects of other sociodemographic factors such as sex, literacy and rural or urban residence on HIV-related knowledge are also explored. In paper 2, a trend analysis is conducted of HIV-related knowledge in the country from 2003 to 2013, with changes in these trends represented graphically, stratified by various sociodemographic factors. ARIMA models were fit to the 2003-2013 trend data. Finally, Paper 3 presents a systematic review (using the Medline and Embase databases) and meta-analysis (conducted in R) of HIV-related knowledge interventions in Sub-Saharan Africa or among the African Diaspora, synthesising the available evidence for the efficacy of such interventions in 1) improving HIV-related knowledge, 2) resulting in increased engagement in preventive measures and safe sexual practices, and 3) reducing HIV incidence. Random-effects models were used for the meta- analyses. Results: The logistic regression model indicated that females were more than twice as likely as males to have low HIV-related knowledge in each wealth inequality category. In addition, females were more likely to have correct knowledge of mother-to-child transmission than males, but were over 1.5 times more likely to have poor knowledge of HIV risk reduction measures. Individuals with lower literacy levels were almost twice as likely as literate respondents to have low HIV-related knowledge. Ethnicity, religious affiliation, relationship status, and residing in rural areas were additional significant predictors of HIV-related knowledge. The trend analysis indicated an overall increase in HIV-related knowledge between 2003 and 2013, but a decrease in knowledge of mother-to-child-transmission. In addition, State-level disparities in knowledge regarding HIV risk reduction increased over time. The meta-analysis of HIV education interventions demonstrated significantly higher odds of correct knowledge of transmission routes as well as condom use, but insignificantly lower odds of HIV incidence. Conclusions: HIV-related knowledge in this sample is generally low among females, those with low literacy levels, the poor, the unemployed, those residing in rural areas, those with traditional religious beliefs, and those living in states with the highest wealth inequality ratios. The meta-analysis of HIV-related knowledge interventions in Paper 3 indicates that such interventions are generally effective at improving not only HIV-related knowledge but also increasing condom use, and should thus be targeted at the risk groups identified in Papers 1 and 2, in order to work towards the reduction of HIV transmission.

Made available in DSpace on 2016-06-02T20:06:12Z (GMT). No. of bitstreams: 1 DissJBB.pdf: 632747 bytes, checksum: 7f6e8caa78736a965ecb167ee27b7cc3 (MD5) Previous issue date: 2006-04-18 / Universidade Federal de Sao Carlos / This work consider the problem of how to incorporate model selection uncertainty into statistical inference, through model averaging, applied to logistic regression. It will be used the approach of Buckland et. al. (1997), that proposed an weighed estimator to a parameter common to all models in study, where the weights are obtained by information criteria or bootstrap method. Also will be applied bayesian model averaging as shown by Hoeting et. al. (1999), where posterior probability is an average of the posterior distributions under each of the models considered, weighted by their posterior model probability. The aim of this work is to study the behavior of the weighed estimator, both, in the classic approach and in the bayesian, in situations that consider the use of binary logistic regression, with foccus in prediction. The known model-choice selection method Stepwise will be considered as form of comparison of the predictive performance in relation to model averaging. / Esta dissertação considera o problema de incorporação da incerteza devido à escolha do modelo na inferência estatística, segundo a abordagem de ponderação de modelos, com aplicação em regressão logística. Será utilizada a abordagem de Buckland et. al. (1997), que propuseram um estimador ponderado para um parâmetro comum a todos os modelos em estudo, sendo que, os pesos desta ponderação são obtidos a partir do uso de critérios de informação ou do método bootstrap. Também será aplicada a ponderação bayesiana de modelos como apresentada por Hoeting et. al. (1999), onde a distribuição a posteriori do parâmetro de interesse é uma média da distribuição a posteriori do parâmetro sob cada modelo em consideração ponderado por suas respectivas probabilidades a posteriori. O objetivo deste trabalho é estudar o comportamento do estimador ponderado, tanto na abordagem clássica como na bayesiana, em situações que consideram o uso de regressão logística binária, com enfoque na estimação da predição. O método de seleção de modelos Stepwise será considerado como forma de comparação da capacidade preditiva em relação ao método de ponderação de modelos.

abstract: Modern, advanced statistical tools from data mining and machine learning have become commonplace in molecular biology in large part because of the “big data” demands of various kinds of “-omics” (e.g., genomics, transcriptomics, metabolomics, etc.). However, in other fields of biology where empirical data sets are conventionally smaller, more traditional statistical methods of inference are still very effective and widely used. Nevertheless, with the decrease in cost of high-performance computing, these fields are starting to employ simulation models to generate insights into questions that have been elusive in the laboratory and field. Although these computational models allow for exquisite control over large numbers of parameters, they also generate data at a qualitatively different scale than most experts in these fields are accustomed to. Thus, more sophisticated methods from big-data statistics have an opportunity to better facilitate the often-forgotten area of bioinformatics that might be called “in-silicomics”. As a case study, this thesis develops methods for the analysis of large amounts of data generated from a simulated ecosystem designed to understand how mammalian biomechanics interact with environmental complexity to modulate the outcomes of predator–prey interactions. These simulations investigate how other biomechanical parameters relating to the agility of animals in predator–prey pairs are better predictors of pursuit outcomes. Traditional modelling techniques such as forward, backward, and stepwise variable selection are initially used to study these data, but the number of parameters and potentially relevant interaction effects render these methods impractical. Consequently, new modelling techniques such as LASSO regularization are used and compared to the traditional techniques in terms of accuracy and computational complexity. Finally, the splitting rules and instances in the leaves of classification trees provide the basis for future simulation with an economical number of additional runs. In general, this thesis shows the increased utility of these sophisticated statistical techniques with simulated ecological data compared to the approaches traditionally used in these fields. These techniques combined with methods from industrial Design of Experiments will help ecologists extract novel insights from simulations that combine habitat complexity, population structure, and biomechanics. / Dissertation/Thesis / Masters Thesis Industrial Engineering 2018

Despite years of deepening economic integration among the states and regions of the European Union, empirical research remains inconclusive about speed of convergence across regions, if not its existence. This thesis provides novel evidence on convergence in the EU while focusing on development at regional level after the Great Recession. It uses recently developed log t convergence test by Phillips and Sul (2007). Our findings speak against the convergence in level of income per capita among the European regions and give us five clubs of regions converging in their income growth rates instead. Investigating further the geographical distribution of the convergence clubs, we confirm high inequality within the member states and find large continuous area of high convergence clubs in the urbanized part of Western Europe. Furthermore, we investigated the determinants of convergence club membership using Logistic Regression. The main explanatory variable of interest were Business Services (BS), a dynamic sector of the economy with presumably strong positive effect on regional innovative potential. We found positive effect of BS on membership in higher convergence clubs. Yet, this effect seems to diminish for the very highest club.