This study utilized a new database of levee failures along 685 km of the Middle and Lower Mississippi River from St. Louis, MO to Memphis, TN during the past 120 years. The goals of this investigation were to: 1) identify the relative importance of geologic and geomorphic factors that have led to levee failures through the past century along the Mississippi River and 2) measure levee crest elevations to determine if they have increased or decreased between 1998 and 2007 and if they are built to the proper design grade elevation. Logistic regression analysis was utilized to examine selected site characteristics at each levee failure location. These site characteristics (levee failure parameters) included: 1) levee underlain by previous channel fill, 2) presence/absence of borrow pit, 3) location of failure on a meander bend, 4) width of channel, 5) width of floodway, 6) constriction-over-time factor, 7) land-cover type, 8) width of vegetative buffer, 9) sinuosity of channel, 10) intensity of dredging, and 11) presence/absence of bank revetment. Each of these parameters was evaluated using geologic maps, soil survey data, Digital Elevation Models (DEMs), historic river maps, and dredging reports (Winkley, 1977; Pinter et al., 2004). Two models were created for each river reach. The first model for both reaches used a 95% significance threshold, while the second model for the MMR used a 80% significance threshold, and the second model for the LMR used a 90% significance threshold. The first model for the Middle Mississippi River (MMR) identified only the presence/absence of channel fills to predict levee failure as significant, had an R² value of 0.178, a p-value of 0.002, and a percentage accuracy of 68.6%. The second model for the MMR identified the following variables as significant: presence/absence of channel fills to predict levee failure, location of failure on a meander bend, channel width, land-cover type, and intensity of dredging. This model had an R2 value of 0.408, p-value of 0.002, and a percentage accuracy of 74.3%. The 95% model for the Lower Mississippi River (LMR) identified location of failure on a meander bend, land-cover type, constriction-over-time factor, and sinuosity of the channel as significant. This model had an R2 value of 0.326, p-value of 0.003, and a percentage accuracy of 69.5%. The 90% LMR model identified the following variables as significant: the presence/absence of borrow pits, location of failure on meander bend, channel width, land-cover type, constriction-over-time factor, vegetative buffer width, channel sinuosity, and presence/absence of bank revetment. This model had an R2 value of 0.385, p-value of 0.006, and a percentage accuracy of 72.0%. The MMR and LMR models with the 95% significance threshold had no predictors in common because of differences between the two river reaches or possibly because of the small sample size. However, the expanded MMR and LMR models shared three predictors (i.e., meander location, channel width, and land cover type). The second portion of this project used post-processed dual-frequency GPS surveying to measure levee elevations between St. Louis, MO and Cairo, IL. These elevations were compared to the 50-year design flood grade elevations and to a 1998 DEM to identify areas of levee heightening, levee degradation and/or subsidence, and locations of past levee crevasses. This surveying revealed areas that, between 1998 and 2007, were raised up to ~1.5 meters and other areas that degraded or subsided up to ~1.0 meter. Also, the locations of five recent levee crevasses were investigated, showing local increases or decreases at those points. The importance of levee road construction type was identified through many sharp increases or decreases at the transition between road types. Overall, this project showed promising implications for the determination of levee failure susceptibility and proper levee elevation heights using logistic regression analysis and kinematic GPS surveying. The logistic regression models predicted the potential for levee failure based on local site characteristics of levees between St. Louis, MO and Memphis, TN. The high-precision kinematic GPS surveying illustrated levee elevations along the MMR to a high degree of accuracy, allowing for the rapid and efficient identification of areas that do not correspond to the proper design flood grade elevation.

Working within the National Longitudinal Transition Study (NLTS) theoretical framework, the purpose of this study was to explore the effects of individual characteristics, in-school experiences, post-school experiences, and contextual factors on Vocational Rehabilitation (VR) closure status among 4,443 young adults with disabilities who had received and completed services from Oregon VR between 2003 and 2013. This study analyzed extant data from the Oregon Rehabilitation Case Automation System (ORCA), an integrated case management database that collects and tracks demographic characteristics, service records and employment data on each individual who receives services from VR. Four logistic regression models were developed using Hosmer, Lemeshow, and Studivant's model building approach to test the effects of individual characteristics, in-school experiences, post-school experiences, and contextual factors on VR closure status. Seven risk factors were identified that decrease the probability of young adults with disabilities achieving a positive VR closure status: (1) being female; (2) having a primary disability of mental illness; (3) having a primary disability of traumatic brain injury; (4) having an interpersonal impediment to employment; (5) receiving Supplemental Security Income at application; (6) closing VR services during federal fiscal year (FFY) 2008; and (7) closing VR services during FFY 2009. Five protective factors were identified that increase the probability of young adults with disabilities achieving a positive VR closure status: (1) participation in the Oregon Youth Transition Program; (2) earning at least a high school completion certificate by closure; (3) receiving a higher number of VR services; (4) closing VR services on or below the median number of days to closure; and (5) closing VR services during FFY 2004. These findings support the hypothesis that individual characteristics, in-school experiences, post-school experiences, and contextual factors are predictors of positive VR closure status among young adults with disabilities. Further, these results provide evidence that transition services and supports provided to young adults with disabilities receiving services from VR can help them to achieve positive VR closure status.

This thesis investigates the factors that are associated with annual changes in the share price of Johannesburg Stock Exchange (JSE) listed companies. In this study, an increase in value of a share is when the share price of a company goes up by the end of the financial year as compared to the previous year. Secondary data that was sourced from McGregor BFA website was used. The data was from 2004 up to 2011. Deciding which share to buy is the biggest challenge faced by both investment companies and individuals when investing on the stock exchange. This thesis uses binary logistic regression to identify the variables that are associated with share price increase. The dependent variable was annual change in share price (ACSP) and the independent variables were assets per capital employed ratio, debt per assets ratio, debt per equity ratio, dividend yield, earnings per share, earnings yield, operating profit margin, price earnings ratio, return on assets, return on equity and return on capital employed. Different variable selection methods were used and it was established that the backward elimination method produced the best model. It was established that the probability of success of a share is higher if the shareholders are anticipating a higher return on capital employed, and high earnings/ share. It was however, noted that the share price is negatively impacted by dividend yield and earnings yield. Since the odds of an increase in share price is higher if there is a higher return on capital employed and high earning per share, investors and investment companies are encouraged to choose companies with high earnings per share and the best returns on capital employed. The final model had a classification rate of 68.3% and the validation sample produced a classification rate of 65.2% / Mathematical Sciences / M.Sc. (Statistics)

Master of Science / Department of Statistics / Trevor Hefley / Binary regression models for spatial data are commonly used in disciplines such as epidemiology and ecology. Many spatially-referenced binary data sets suffer from location error, which occurs when the recorded location of an observation differs from its true location. When location error occurs, values of the covariates associated with the true spatial locations of the observations cannot be obtained. We show how a change of support (COS) can be applied to regression models for binary data to provide bias-corrected coefficient estimates when the true values of the covariates are unavailable, but the unknown location of the observations are contained within non-overlapping polygons of any geometry. The COS accommodates spatial and non-spatial covariates and preserves the convenient interpretation of methods such as logistic and probit regression. Using a simulation experiment, we compare binary regression models with a COS to naive approaches that ignore location error. We illustrate the flexibility of the COS by modeling individual-level disease risk in a population using a binary data set where the location of the observations are unknown, but contained within administrative units. Our simulation experiment and data illustration corroborate that conventional regression models for binary data which ignore location error are unreliable, but that the COS can be used to eliminate bias while preserving model choice.

abstract: Correlation is common in many types of data, including those collected through longitudinal studies or in a hierarchical structure. In the case of clustering, or repeated measurements, there is inherent correlation between observations within the same group, or between observations obtained on the same subject. Longitudinal studies also introduce association between the covariates and the outcomes across time. When multiple outcomes are of interest, association may exist between the various models. These correlations can lead to issues in model fitting and inference if not properly accounted for. This dissertation presents three papers discussing appropriate methods to properly consider different types of association. The first paper introduces an ANOVA based measure of intraclass correlation for three level hierarchical data with binary outcomes, and corresponding properties. This measure is useful for evaluating when the correlation due to clustering warrants a more complex model. This measure is used to investigate AIDS knowledge in a clustered study conducted in Bangladesh. The second paper develops the Partitioned generalized method of moments (Partitioned GMM) model for longitudinal studies. This model utilizes valid moment conditions to separately estimate the varying effects of each time-dependent covariate on the outcome over time using multiple coefficients. The model is fit to data from the National Longitudinal Study of Adolescent to Adult Health (Add Health) to investigate risk factors of childhood obesity. In the third paper, the Partitioned GMM model is extended to jointly estimate regression models for multiple outcomes of interest. Thus, this approach takes into account both the correlation between the multivariate outcomes, as well as the correlation due to time-dependency in longitudinal studies. The model utilizes an expanded weight matrix and objective function composed of valid moment conditions to simultaneously estimate optimal regression coefficients. This approach is applied to Add Health data to simultaneously study drivers of outcomes including smoking, social alcohol usage, and obesity in children. / Dissertation/Thesis / Doctoral Dissertation Statistics 2018

A inclusão da eletroencefalografia nos protocolos de pesquisa diagnóstica para DA é plenamente justificada por sua larga disponibilidade, baixo custo, alta sensibilidade, o que possibilita a realização de exames seriados e o acompanhamento da evolução do estudo neurológico. Objetivo: Determinar um índice de corte, para utilizaçào na prática clínica, no auxilio diagnóstico da doença de Alzheimer. Metodologia: Avaliamos dois grupos de indivíduos compostos por 57 voluntários normais e idade superior a 50 anos comparados a 50 indivíduos com DA provável. Realizamos registros de EEG em vigília, olhos fechados e repouso por 30 minutos e computamos as potências espectrais das bandas de frequência alfa e teta, para todos os eletrodos, e calculamos a razão alfa/teta. Realizamos a regressão logística das variáveis razão alfa/teta da potência média do eletrodo C3 e do eletrodo O1e calculamos uma fórmula para o auxílio no diagnóstico da DA com um acerto cuja, sensibilidade para DA de 76, 4 % e especificidadede 84,6 % e a área sob a curva ROC 0.92. Conclusão: A regressão logística da razão alfa/teta do Espectro da potência média do EEG é um bom marcador para discriminar pacientes com doença de Alzheimer de controles normais / The inclusion of electroencephalography in diagnostic research protocols for AD is fully justified given EEG\'s wide availability, low cost and high sensitivity, allowing serial exams and neurological evolution follow-ups. Objective: To determine a screening index for use in routine clinical practice to aid the diagnosis of Alzheimer\'s disease. Methodology: Two groups of individuals older than 50 years, comprising a control group of 57 normal volunteers and a study group of 50 patients with probable AD, were compared. EEG recordings were performed of subjects in a wake state with eyes closed at rest for 30 mins. Spectral potentials of the alpha and theta bands were computed for all electrodes and the alpha/theta ratio calculated. Logistic regression of the variables alpha/theta of the mean potential of the C3 and O1 electrodes was carried out. A formula was calculated to aid the diagnosis of AD yielding 76.4 % sensitivity and 84.6 specificity for AD with an area under the ROC curve of 0.92. Conclusion: Logistic regression of the alpha/theta of the Spectrum of the mean potential of EEG represents a good marker for discriminating between AD patients and normal controls

As a consequence from the last financial crisis that began 2007 in USA, regulatory frameworks are continuously improved in order to limit the banks’ risk exposure. Two of the amendments are Basel III and IFRS 9. Basel III regulates the capital a bank is required to hold while IFRS 9 is an accounting standard for how banks and insurance companies should classify their assets and estimate their future credit losses. Mutually for both Basel III and IFRS 9 is the estimation of future credit losses which include probability of default in the calculations.The objective of this thesis was therefore to develop scoring model that can estimate the probability of default in lending capital to enterprises based on information from financial statements. The aim is that the developed model also can be used in the daily operations to reduce fixed costs by optimizing the processes and increase the profit on each loan issued. The model should estimate probability of default within 500 days from the last known information and be customized for small and medium size enterprises.The model is based on logistic regression and is therefore returning values between 0 and 1. Parameters that the model consists of can either be calculated or retrieved directly from financial statements. The authors have during the development of the model divided the data, consisting of information from enterprises, based on branches. The grouping of data has been performed to create as homogenous sets of data as possible in order to increase the degree of explanation for each model. The final solution will thus consist of several models, one for each set of data. The validation of the models is performed, on a new set of enterprises where it is observed how well the models can discriminate enterprises defined as defaults from non-defaults.The master thesis did result in a number of models that are calibrated on default, non-defaults and models developed on data divided on branches. By using the calibrated models, it is possible to discriminate defaulting from non-defaulting enterprises which has been the objective of this thesis. During the project the importance of dividing data into homogenous groups has been shown in order to better create models that more accurately can identify defaults from non-defaults. / Som en konsekvens av finanskrisen som började 2007 i USA tillkom ytterligare regelverk för att minimera bankers riskexponering. Två av de regelverk som tillkommit är Basel III och IFRS 9. Basel III reglerar kapitaltäckningen för en bank medan IFRS 9 är en standard för hur banker och försäkringsbolag skall klassificera tillgångar samt estimera framtida kreditförluster. Gemensamt för de båda regelverken är estimeringen av kreditförluster som bland annat baseras på risken för fallissemang.Målet med detta examensarbete är därför att utveckla en scoringmodell som kan estimera risken för fallissemang vid utlåning till företag baserat på information från dess årsredovisningar. Modellen kommer även kunna användas i den operativa verksamheten för att reducera fasta kostnaderna genom att effektivisera processer och då öka avkastningen på varje utlånad krona. Modellen kommer att estimera risken för fallissemang inom 500 dagar från senast kända informationen och den kommer att anpassas till svenska små och medelstora företag.Modellen är baserad på logistisk regression och kommer därför att returnera värden mellan 0 och 1 samt bestå av parametrar som antingen kan beräknas eller hämtas direkt ur en årsredovisning. För att öka modellens förklaringsgrad har författarna vid kalibreringen av modellerna delat in datat efter branscher. Uppdelningen har gjorts för att skapa så homogena grupper som möjligt och lösningen kommer därför att bestå av flera olika modeller. Validering av modellerna sker genom att på nytt data testa hur bra företag som definierats som fallissemang kan diskrimineras från företag som inte definieras som fallissemang.Rapporten resulterar i ett antal modeller som är baserade på konkurser, icke konkurser samt modeller baserade på ett data som är uppdelat på branscher. Genom att använda de kalibrerade modellerna så går det att diskriminera konkurser från icke konkurser vilket varit målet med denna rapport. Arbetet har också påvisat vikten av att dela in datat i homogena grupper för att på ett bättre sätt skapa modeller som mer exakt kan urskilja konkurser från icke konkurser.

Doctor of Philosophy / Department of Computer Science / William H. Hsu / The contribution of the proposed dissertation is the application of affective intelligence in human-developed spaces where people live, work, and recreate daily, also known as built environments. Built environments have been known to influence and impact individual affective responses. The implications of built environments on human well-being and mental health necessitate the need to develop new metrics to measure and detect how humans respond subjectively in built environments. Detection of arousal in built environments given biometric data and environmental characteristics via a machine learning-centric approach provides a novel and new capability to measure human responses to built environments. Work was also conducted on experimental design methodologies for multiple sensor fusion and detection of affect in built environments. These contributions include exploring new methodologies in applying supervised machine learning algorithms, such as logistic regression, random forests, and artificial neural networks, in the detection of arousal in built environments. Results have shown a machine learning approach can not only be used to detect arousal in built environments but also for the construction of novel explanatory models of the data.

The secondary structure of proteins is predicted using various binary classifiers. The data are adopted from the RS126 database. The original data consists of protein primary and secondary structure sequences. The original data is encoded using alphabetic letters. These data are encoded into unary vectors comprising ones and zeros only. Different binary classifiers, namely the naive Bayes, logistic regression and classification trees using hold-out and 5-fold cross validation are trained using the encoded data. For each of the classifiers three classification tasks are considered, namely helix against not helix (H/∼H), sheet against not sheet (S/∼S) and coil against not coil (C/∼C). The performance of these binary classifiers are compared using the overall accuracy in predicting the protein secondary structure for various window sizes. Our result indicate that hold-out cross validation achieved higher accuracy than 5-fold cross validation. The Naive Bayes classifier, using 5-fold cross validation achieved, the lowest accuracy for predicting helix against not helix. The classification tree classifiers, using 5-fold cross validation, achieved the lowest accuracies for both coil against not coil and sheet against not sheet classifications. The logistic regression classier accuracy is dependent on the window size; there is a positive relationship between the accuracy and window size. The logistic regression classier approach achieved the highest accuracy when compared to the classification tree and Naive Bayes classifiers for each classification task; predicting helix against not helix with accuracy 77.74 percent, for sheet against not sheet with accuracy 81.22 percent and for coil against not coil with accuracy 73.39 percent. It is noted that it is easier to compare classifiers if the classification process could be completely facilitated in R. Alternatively, it would be easier to assess these logistic regression classifiers if SPSS had a function to determine the accuracy of the logistic regression classifier.

Quand vient le temps de comprendre le comportement d’une personne, le regard est une source d’information importante. L’analyse des comportements des consommateurs, des criminels, ou encore de certains états cognitifs passe par l’interprétation du regard dans une scène à travers le temps. Il existe un besoin réel d’identification de la zone regardée sur un écran ou tout autre médium par un utilisateur. Pour cela, la vision humaine fait la composition de plusieurs images pour permettre de comprendre la relation tridimensionnelle qui existe entre les objets et la scène. La perception 3D d’une scène réelle passe alors à travers plusieurs images. Mais qu’en est-il lorsqu’il n’y a qu’une seule image ?