Dissertations / Theses on the topic 'Stepwise regression models'

Made available in DSpace on 2017-07-10T17:48:29Z (GMT). No. of bitstreams: 1 Bianchecci_Danielli_Conte.PDF: 1308564 bytes, checksum: 046371656b2367e67b20775d20072c2b (MD5) Previous issue date: 2011-12-09<br>Fundação Araucária<br>This study aimed to predict the fillet weight (PF) and fillet yield (RF) of Nile tilapia (Oreochromis niloticus) from external measures, measured by morphometry and measures of epiaxial muscles using ultrasound and validate the estimated equations using another sample biometric data. A total of 102 adult fish GIFT lineage, reversed males weighing between 260 and 580 g had the total weight (PT) and body circumference average (CR) all measured by ultrasonography in four predefined body regions (1) between the basis of anterior (2) between the basis of anteriour insertion of anal fin to the last hard ray of dorsal fin. (3) between the final insertion of anal and dorsal fins and (4) between the ventral and dorsal insertion of caudal fin. The prediction equations from the measurements performed on 50 samples were analyzed by statistical procedures backward and stepwise. The measurements made in remaining 52 samples were used for validation of predictive equations to PF and RF. The validation were made from adjustment of linear models first degree of values observed above the values predict on each regression equation set. The equations PFx and RFx estimated from data collected of ultrasound images showed coefficient of determination (R2) of 0.53 (PF1 = -17.35 + 3.66HD2 45.20LE4 + 0.55AE2 + 4.64AE4) e 0.15 (RF1 = 26.66 + 0.043AE2), in backward, and 0.53 (PF2 = -58.03 + 5.65HE4 + 0.33AD1 + 0.65AD2) e 0.19 (RF2 = 25.02 + 0.051AD2), in stepwise. The inclusion of measures of CR and PT as regressive equations did allow the equations to show higher values of R2 such as 0.97 (PF3= - 52.25 1.26HE1 2.03HE3 + 0.40PT +0.51CR e PF4 = -18.12 1.20HE1 + 0.425PT) and 0.73 (RF3 = 17.83 0.29HE1 0.49HE3 + 0.02PT + 0.13CR), in backward and stepwise, and 0.68 (RF4 = 26,14 - 0,27HE1 + 0,027PT), in stepwise, and lower values of mean square of residue that these regression models presented high adherence to data of PF and RF. About the validation of prediction equations all of it were effective in estimating the PF and RF in the sample evaluated. The equations (P = 1.0000) PF3, PF4 (P = 0.5401), RF3 (P = 1.0000) and RF4 (P = 0.8363) were the most accurate and applicable in predicting PF and RF in Nile tilapia. The height of the left side body regions 1 and 3, measured by ultrasound images, together with the average body weight and circumference, are important in predicting of regressive of PF and RF of Nile tilapia. The regression equations, PF3, PF4, RF3, and RF4 are recommended for estimating the PF and RF of Tilapia in phenotypic males adult Nile tilapia GIFT lineage, weighing between 260 and 580 g<br>Este estudo teve por objetivos predizer o peso de filé (PF) e o rendimento de filé (RF) de tilápia do Nilo (Oreochromis niloticus), a partir de medidas externas mensuradas por morfometria e medidas da musculatura epiaxial, mensuradas por ultrassonografia, e validar as equações estimadas utilizando outra amostra de dados biométricos. Um total de 102 peixes adultos, da linhagem GIFT, machos invertidos, pesando entre 260 e 580 g, foram avaliados quanto ao peso total (PT), circunferência corporal média (CR) e mensurados por ultrassonografia em quatro regiões corpóreas pré-definidas: (1) entre a base de inserção anterior da nadadeira pélvica até o término anterior da nadadeira dorsal, (2) entre a base da inserção anterior da nadadeira anal até o último raio duro da nadadeira dorsal, (3) entre a inserção final das nadadeiras anal e dorsal e (4) entre a inserção ventral e dorsal da nadadeira caudal. As equações de predição a partir das mensurações realizadas em 50 exemplares foram analisadas pelos procedimentos estatísticos backward e stepwise. As mensurações feitas nos 52 exemplares restantes foram utilizadas para a validação das equações preditoras para PF e RF. A validação foi feita a partir do ajuste de modelos lineares de 1° grau dos valores observados sobre os valores preditos por cada equação de regressão definida. As equações PFx e RFx, estimadas em função dos dados coletados das imagens ultrassonográficas, apresentaram valores de coeficiente de determinação (R²) de 0,53 (PF1 = -17,35 + 3,66HD2 - 45,20LE4 + 0,55AE2 + 4,64AE4) e 0,15 (RF1 = 26,66 + 0,043AE2), no backward, e 0,53 (PF2 = -58,03 + 5,65HE4 + 0,33AD1 + 0,65AD2) e 0,19 (RF2 = 25,02 + 0,051AD2), no stepwise. A inclusão das medidas de CR e PT como regressoras permitiu que as equações apresentassem R² mais elevados, com valores de 0,97 (PF3 = - 52,25 - 1,26HE1 - 2,03HE3 + 0,40PT +0,51CR e PF4 = -18,12 - 1,20HE1 + 0,425PT) e 0,73 (RF3 = 17,83 - 0,29HE1 - 0,49HE3 + 0,02PT + 0,13CR), no backward e no stepwise, e 0,68 (RF4 = 26,14 - 0,27HE1 + 0,027PT), no stepwise, e menores valores de quadrado médio do resíduo, indicando que esses modelos de regressão apresentaram elevada aderência aos dados de PF e RF. Na validação das equações preditoras, todas foram eficientes em estimar o PF e o RF na amostra avaliada. As equações PF3 (P = 1,0000), PF4 (P = 0,5401), RF3 (P = 1,0000) e RF4 (P = 0,8363) foram as mais precisas e aplicáveis na predição de PF e RF em tilápia do Nilo. A altura do lado esquerdo nas regiões corporais 1 e 3, mensuradas por imagens ultrassonográficas, em conjunto com o peso e a circunferência corporal média, são regressoras importantes na predição do PF e RF de tilápia do Nilo. As equações de regressão PF3, PF4, RF3 e RF4 são recomendadas para estimar o PF e o RF de tilápia no Nilo adultas, machos fenotípicos, linhagem GIFT, com peso entre 260 e 580g

O padrão espacial de ocorrência de atributos de espécies florestais, tal como a área basal das árvores, pode fornecer informações para o entendimento da estrutura da comunidade vegetal. Uma vez que fatores ambientais podem influenciar tanto o padrão espacial de ocorrência quanto os atributos das espécies em florestas nativas. Desse modo, investigar a relação entre as características ambientais e o padrão espacial de espécies florestais pode ajudar a entender a dinâmica das florestas. Especificamente, neste trabalho, o objetivo é avaliar métodos estatísticos que permitam identificar quais atributos do solo são capazes de explicar a variação da área basal de cada espécie de árvore. A área basal foi considerada como variável resposta e como covariáveis, um grande número de atributos físicos e químicos do solo, medidos em uma malha de localizações cobrindo a área de estudo. Foram revisados e utilizados os métodos de regressão linear múltipla com método de seleção stepwise, modelos aditivos generalizados e árvores de regressão. Em uma segunda fase das análises, adicionou-se um efeito espacial aos modelos, com o intuito de verificar se havia ainda padrões na variabilidade, não capturados pelos modelos. Com isso, foram considerados os modelos autoregressivo simultâneo, condicional autoregressivo e geoestatístico. Dado o grande número de atributos do solo, as análises foram também conduzidas utilizando-se as covariáveis originais, fatores identificados em uma análise fatorial prévia dos atributos de solo. A seleção de modelos com melhor ajuste foi utilizada para identificar os atributos de solo relevantes, bem como a presença e melhor descrição de padrões espaciais. A área de estudo foi a Estação Ecológica de Assis, da Unidade de Conservação do Estado de São Paulo em parcelas permanentes, dentro do projeto \"Diversidade, Dinâmica e Conservação em Florestas do Estado de São Paulo: 40 ha de parcelas permanentes\", do programa Biota da FAPESP. As análises reportadas aqui se referem à área basal das espécies Copaifera langsdorffii, Vochysia tucanorum e Xylopia aromatica. Com os atributos de solo reduzidos e consistentemente associados à área basal, a declividade, altitude, saturação por alumínio e potássio mostraram-se relevantes para duas das espécies. Resultados obtidos mostraram a presença de um padrão na variabilidade, mesmo levando-se em consideração os efeitos das covariáveis, ou seja, os atributos do solo explicam parcialmente a variabilidade da área basal, mas existe um padrão que ocorre no espaço que não é capturado por essas covariáveis.<br>The spatial pattern of occurrenceis of forest species and their attributes, such as the basal area of trees, can provide information for understanding the structure of the vegetable community. Considering the environmental factors can influence the spatial pattern of occurrences of species in native forests and related attributes, describing relationship between environmental characteristics and spatial pattern of forest species can be associated with the dynamics of forests. The objective of the present study is to assess different statistical methods used to identify which soil attributes are associated with the basal area of each tree selected species. The basal area was considered as the response variable and the covariates are given by a large number of physical and chemical attributes of the soil, measured at a grid of locations covering the study area. The methods considered are the multiple linear regression with stepwise model selection, generalized additive models and regression trees. Spatial effects were added to the models, in order to ascertain whether there is residual spatial patterns not captured by the covariates. Thus, simultaneous autoregressive model, autoregressive conditional and geostatistical were considered. Considering the large number of soil attributes, analysis were were conducted both ways, using the original covariates, and using factors identified in a preliminar factor analysis of the soil attributes. Model selection was used to identify the relevant attributes of soil as well as the presence and better description of spatial patterns. The study area was the Ecological Station of Assis, the Conservation Unit of the State of São Paulo in permanent plots within the \"Diversity Dynamics and Conservation Forests in the State of São Paulo: 40 ha of permanent plots\" project, under the research project FAPESP biota. The analyzes reported here refer to the basal area of the species Copaifera langsdorffii, Vochysia tucanorum and Xylopia aromatica. Results differ among the considered methods reinforcing the reccomendation of considering differing modeling strategies. Covariates consistently associated with basal area are slope, altitude and aluminum saturation, potassium, relevant to at least two of the species. Results obtained showed the presence of patterns in residual variability, even taking into account the effects of covariates. The soil characteristics only partially explain the variability of the basal area and there are spatial patterns not captured by these covariates.

Optimisation of the conventional tripartite of pancreatic cancer (PC) treatment have led to significant improvements in mortality, however further knowledge of the underlying molecular processes is still required. Transcript profiling of mRNA expression of over 44K genes with microarray technology demonstrated upregulation of secreted frizzled related protein 4 (sFRP4) and ??-catenin in PC compared to normal pancreata. Their pathway ??? Wnt signalling is integral to transcriptional regulation and aberrations in these molecules are critical in the development of many human malignancies. Immunohistochemistry protocols were evaluated by two independent blinded examiners for antigen expression differences associated with survival patterns in 140 patients with biopsy verified PC and a subset of 23 normal pancreata with substantial observer agreement (kappa value 0.6-0.8). A retrospective cohort was identified from 6 Sydney hospitals between 1972-2003 and archival formalin fixed tissue was collected together with clinicopathological data. Three manual stepwise regression models were fitted for overall, disease-specific and relapse-free survival to determine the value of significant prognostic variables in risk stratification. The models were fitted in a logical order using a careful strategy with step by step interpretation of the results. Immunohistochemistry demonstrated increased sFRP4 membranous expression (&gt 10%) in 49/95 PC specimens and this correlated with improved overall survival (HR:0.99;95%CI:0.97-6.40;LRchi2=134.75; 1df; ??&lt 0.001). Increased sFRP4 cytoplasmic staining (&gt 2/3) in 46/85 patients increased the disease-specific survival (HR:0.52;95%CI:0.31-0.89;LR test statistic =248.40;1df;??&lt 0.001). Increasing ??-catenin membranous expression (&lt _60%) in 26/116 patients was associated with an increased risk of overall death (HR:3.18;95%CI:1.14-8.89;LR test statistic =4.61;1df,??&lt 0.05). Increasing cytoplasmic expression in 65/114 patients was protective and was associated with prolonged survival on univariate, but not multivariate analysis (Disease specific survival HR:0.75;95%CI:0.56-1.00;logrank chi2=3.91;1df; ??=0.05). Increased nuclear ??-catenin expression in 65/114 patients was associated with prolonged survival (disease-specific HR:0.92;95%CI:0.83-1.02; LR test statistic= 49.72;1df;??&lt 0.001). At the conclusion, 12 patients (8.6%) remained alive, 122 died of their disease (68 males versus 54 females). They were followed for a median of 8.7 months (range 1.0-131.3) months. The median age was 66.5 years (range 34.4-96.0, standard deviation 10.9) years. Pancreatic resection was achieved in 79 patients with 46.8% achieving RO resection. The 30 day post-operative mortality was 2.1%. The overall 1 year survival rate was (33.7% ; 95%CI: 25.78-33.79) with a 5 year survival of (2.87%, 95%CI: 2.83-6.01) and a median survival of (8.90 months; 95%CI: 7.5-10.2). The median disease-specific survival was (9.40; 95%CI: 7.9-10.5 months) and the median time to relapse was 1.2 months (95%CI 1.0-1.2 months). A central tenet of contemporary cancer research is that an understanding of the genetic and molecular abnormalities that accompany the development and progression of cancer is critical to further advances in diagnosis, treatment and eventual prevention. High throughput tissue microarrays were used to study expression of two novel tumour markers in a cohort of pancreatic cancer patients and identified sFRP4 and ??-catenin as potential novel prognostic markers.

Construction projects in developing countries may suffer from time overruns, which are associated with cost overruns. This research project investigated both time and cost overruns on high-rise building projects in Dhaka, Bangladesh. Surprisingly, preliminary data analysis showed negligible cost overruns in comparison to time overruns. So, further analysis o f cost overruns was not considered in this thesis. This research project also investigated how the causes o f time-overruns can be mitigated. 72 time-overrun and 22 cost-overrun variables were identified through a literature review. These variables were taken as parameters and a personal interview survey was conducted with developers, consultants, contractors and project managers using semistructured questionnaire. A similar second survey was conducted using 22 measures, which can mitigate time- overruns. Data analysis involved the relative importance index to rank the variables, factors analysis to reduce variables to factors with minimum loss of data, stepwise regression to find links among factors in successive stages of construction process and multiple regression to explain delays in terms of factors. The main causes o f time-overruns were ‘cash flow’, ‘planning and scheduling deficiency’ and ‘design changes’. A scree graph identified 31 important variables that caused delays but factor analysis reduced these to 14 factors. Stepwise regression found no strong links among the factors to identify them as reasons for delay in successive stages of the construction project. A multiple regression model explained about 85% of the variance of the delays using eight factors. The main individual measures mitigating time-overruns were ‘improvement of cash flow’, ‘improvement o f communication and coordination among project participants’ and ‘development o f robust planning and scheduling instruments’. Factor analysis produced ten representative factors. Stepwise regression could not find strong links among factors mitigating time-overruns in successive stages of the construction project.

Doctor of Philosophy<br>Department of Human Nutrition<br>Edgar Chambers IV<br>Companies introduce new products with the goal of achieving success. However, many products fail. The overall objective of this research was to design processes for determining sensory and market characteristics of food products that could predict success. The first sub-objective was to determine if success could be predicted using information known before launch. The second sub-objective was to describe a process for determining specific sensory characteristics that promote success. Most methods chosen for this research are commonly used. However, previous research has identified a relationship between consumers liking and salivation, without defining a method. Thus, three salivation methods were selected for initial testing: spit, cotton rolls and sensory scale. These were tested on foods with different textures. Although all methods gave similar results, the spit method was chosen for further testing of flavor differences. Differences in salivation measurements were found for snacks where flavors were different but texture was unchanged. Next, flavored snack products from 15 countries were selected that were successful or had failed. Questionnaires were completed for each product and included questions related to authenticity, familiarity, current trends, packaging and marketplace issues such as product competition and pricing, all of which would be known before launch. A discriminant function was developed that correctly identified 75.8% of the successful flavored snack products as successful and 66.7% of the unsuccessful products as unsuccessful. Stepwise comparisons were used to determine that four variables are necessary to correctly categorize these products. The products then were clustered into three groups to select 34 products from 11 countries for further sensory testing. Information from extensive sensory descriptive methods were evaluated individually and in various combinations through stepwise regression and discriminant analysis. The final sensory model correctly predicted all successful and unsuccessful products, had an R-square of 0.84 and included nine regression factors: seven flavor attributes and two flavor attribute ratios. Many of the attributes were base flavor notes necessary for this flavored snack category. A process for selecting key attributes for success was described. For this snack category, creating products with flavors that interact well with base flavor notes can lead to a successful product.

The dissertation focuses on the different determinants of education, their effects on the educational outcome, and the overall effect of education on the lifetime consequences. The first chapter focuses on the inequality of educational opportunity across different demographic factors. This chapter employs a broader set of social factors to provide fresh insights into the inequality situation in the USA relative to those of the extant literature. The chapter employs polynomial trends for the effects of social factors to identify long-term trends in the determinants of the differences in attainment of each of four achievements (high school graduation, some college, college graduation, and post-college work) across different endogenous social groups. Using the Panel Study of Income Dynamics (PSID) data for the years of 1968-2013, we show how inequality of educational opportunity and its determinants have evolved over the years. The chapter utilizes the machine-learning process and logistic regression model to identify inequality of opportunity. The second chapter examines the age demographic distribution of graduates across cohorts from 1940 until 1990. Using the PSID data, the paper explored the first and second moment of the age of graduating from high school and college across the US. To deal with the data deficiencies, a large part of the chapter dealt with data preparation. The chapter provides a unique method of extracting information on the graduating age of the individuals both from high school and from college. The results show a large dispersion across the full sample. The data truncated to a standard length, however, provides a much smaller dispersion and much smaller moments. The chapter concludes that as the time passes, people tend to attain education at a younger age. The third chapter investigates the trends of the contribution of different factors of income starting from 1910 cohort. Following Mincer (1974), a wave of papers studied how various factors contribute to the earnings of individuals. This paper contributes to that literature in three ways: (i) using the PSID data, it computes the actual working experience of the individuals, (ii) it studies the cohorts who were born in 1910 or afterwards, unlike the existing papers, and (iii) it adds two variables�"technological progress and the occupation with which individuals start their careers�"to an extended Mincerian equation. The results re-emphasize the importance of education in lifetime earnings. The results also show that while some of the determinants of income have become more important over the years, other factors have not changed much in importance.<br>PHD

This doctoral study aims to develop a methodology for use in determining aerodynamic models and parameters from actual flight test data for different types of autonomous flight vehicles. The stepwise regression method and equation error method are utilized for the aerodynamic model identification and parameter estimation. A closed loop aerodynamic parameter estimation approach is also applied in this study which can be used to fine tune the model parameters. Genetic algorithm is used as the optimization kernel for this purpose. In the optimization scheme, an input error cost function is used together with a final position penalty as opposed to widely utilized output error cost function. Available methods in the literature are developed for and mostly applied to the aerodynamic system identification problem of piloted aircraft<br>a very limited number of studies on autonomous vehicles are available in the open literature. This doctoral study shows the applicability of the existing methods to aerodynamic model identification and parameter estimation problem of autonomous vehicles. Also practical considerations for the application of model structure determination methods to autonomous vehicles are not well defined in the literature and this study serves as a guide to these considerations.

碩士<br>淡江大學<br>土木工程學系碩士班<br>102<br>Multivariate regression analysis is usually employed to establish real estate valuation formula. This approach is quite simple, but it has a few drawbacks including being difficult to understand the meaning of the coefficients, cannot be universal to another areas, cannot be used in comparison approach, and without good flexibility. The purpose of this study is to propose the stepwise decomposition regression analysis to overcome these shortcomings. The factors considered in this study include The factor of the distance to the nearest MRT station which represents the impact of transportation function to the price per unit area. The factor of the number of convenience stores in the living circle on foot which represents the impact of living function to the price per unit area. The factor of the age of house which represents the impact of the quality of the house to the price per unit area. The factor of transaction date which represents the impact of market trend to the price per unit area. The factor of the geographic coordinates which represent the impact of spatial location to the price per unit area. The results showed that the 20% error hit rates of real estate valuation were greater than 70% for all the four testing areas in Taipei City and New Taipei City.

碩士<br>淡江大學<br>土木工程學系碩士班<br>103<br>The purpose of this study is to propose the stepwise decomposition regression analysis, quantitative comparative approach, and their mixed method to build real estate valuation models. The valuation target is the price per unit area. The factors considered in this study include distance to the nearest MRT station, amount of convenience stores in the living circle, age of the house, transaction date, geographic coordinates and width of the road. The study areas are all the twelve administrative areas of Taipei City. The results showed that (1) the distance to the nearest MRT station is the most important factor among all the factors. (2) Distance to the nearest MRT station and age of house are inversely proportional factors; amount of convenience stores in the living circle and transaction date are proportional factors. The analysis of geographic coordinates showed that the closer to the center of Taipei, the higher the real estate price. But the effect of width of road is not significant. (2) The 20% error hit rates of real estate valuation of quantitative comparative approach were 5% greater than the stepwise decomposition regression analysis. (3) The mixed methods are more accurate than the quantitative comparative approach but not significantly.

Return on Capital Employed (ROCE) is a popular financial instrument and communication tool for the appraisal of companies. Often, companies management and other practitioners use untested rules and behavioural approach when investigating the key determinants of ROCE, instead of the scientific statistical paradigm. The aim of this dissertation was to identify and quantify key determinants of ROCE of individual companies listed on the Johannesburg Stock Exchange (JSE), by comparing classical multiple linear regression, principal components regression, generalized least squares regression, and robust maximum likelihood regression approaches in order to improve companies decision making. Performance indicators used to arrive at the best approach were coefficient of determination ( ), adjusted ( , and Mean Square Residual (MSE). Since the ROCE variable had positive and negative values two separate analyses were done. The classical multiple linear regression models were constructed using stepwise directed search for dependent variable log ROCE for the two data sets. Assumptions were satisfied and problem of multicollinearity was addressed. For the positive ROCE data set, the classical multiple linear regression model had a of 0.928, an of 0.927, a MSE of 0.013, and the lead key determinant was Return on Equity (ROE),with positive elasticity, followed by Debt to Equity (D/E) and Capital Employed (CE), both with negative elasticities. The model showed good validation performance. For the negative ROCE data set, the classical multiple linear regression model had a of 0.666, an of 0.652, a MSE of 0.149, and the lead key determinant was Assets per Capital Employed (APCE) with positive effect, followed by Return on Assets (ROA) and Market Capitalization (MC), both with negative effects. The model showed poor validation performance. The results indicated more and less precision than those found by previous studies. This suggested that the key determinants are also important sources of variability in ROCE of individual companies that management need to work with. To handle the problem of multicollinearity in the data, principal components were selected using Kaiser-Guttman criterion. The principal components regression model was constructed using dependent variable log ROCE for the two data sets. Assumptions were satisfied. For the positive ROCE data set, the principal components regression model had a of 0.929, an of 0.929, a MSE of 0.069, and the lead key determinant was PC4 (log ROA, log ROE, log Operating Profit Margin (OPM)) and followed by PC2 (log Earnings Yield (EY), log Price to Earnings (P/E)), both with positive effects. The model resulted in a satisfactory validation performance. For the negative ROCE data set, the principal components regression model had a of 0.544, an of 0.532, a MSE of 0.167, and the lead key determinant was PC3 (ROA, EY, APCE) and followed by PC1 (MC, CE), both with negative effects. The model indicated an accurate validation performance. The results showed that the use of principal components as independent variables did not improve classical multiple linear regression model prediction in our data. This implied that the key determinants are less important sources of variability in ROCE of individual companies that management need to work with. Generalized least square regression was used to assess heteroscedasticity and dependences in the data. It was constructed using stepwise directed search for dependent variable ROCE for the two data sets. For the positive ROCE data set, the weighted generalized least squares regression model had a of 0.920, an of 0.919, a MSE of 0.044, and the lead key determinant was ROE with positive effect, followed by D/E with negative effect, Dividend Yield (DY) with positive effect and lastly CE with negative effect. The model indicated an accurate validation performance. For the negative ROCE data set, the weighted generalized least squares regression model had a of 0.559, an of 0.548, a MSE of 57.125, and the lead key determinant was APCE and followed by ROA, both with positive effects.The model showed a weak validation performance. The results suggested that the key determinants are less important sources of variability in ROCE of individual companies that management need to work with. Robust maximum likelihood regression was employed to handle the problem of contamination in the data. It was constructed using stepwise directed search for dependent variable ROCE for the two data sets. For the positive ROCE data set, the robust maximum likelihood regression model had a of 0.998, an of 0.997, a MSE of 6.739, and the lead key determinant was ROE with positive effect, followed by DY and lastly D/E, both with negative effects. The model showed a strong validation performance. For the negative ROCE data set, the robust maximum likelihood regression model had a of 0.990, an of 0.984, a MSE of 98.883, and the lead key determinant was APCE with positive effect and followed by ROA with negative effect. The model also showed a strong validation performance. The results reflected that the key determinants are major sources of variability in ROCE of individual companies that management need to work with. Overall, the findings showed that the use of robust maximum likelihood regression provided more precise results compared to those obtained using the three competing approaches, because it is more consistent, sufficient and efficient; has a higher breakdown point and no conditions. Companies management can establish and control proper marketing strategies using the key determinants, and results of these strategies can see an improvement in ROCE.<br>Mathematical Sciences<br>M. Sc. (Statistics)

碩士<br>中華大學<br>資訊管理學系(所)<br>98<br>After combining many effects, we can construct a stock selection model with higher return, but it is clearly inefficient using trial and error method to find the best multi-factor model. This study took Sorting Normalization Method to normalize the independent variables and the dependent variable. On the modeling method, since only some factors influence the stock selection, an over-complex multi-factor model will fall into the trap of data snooping and will only build a repetition model, not a generalization model. Therefore, the study employed the stepwise regression and regression tree to build stock selection models. To evaluate the accuracy of these models, the study took an empirical analysis on the Taiwan stock market. Conclusions are given as follows. (1) From the 12 factors considered, stepwise regression selected two most important factors, the return on equity (ROE) and price-to-book value ratio (PBR). These two factors could explain more than 80% variance. (2) According to the estimation of stepwise regression, the best weight of ROE and the best weight of PBR are respectively 55% and 45%. Empirical results showed that the weights are close to the best one. (3) Regression trees showed that after-tax earnings per share, PE ratio, and PBR are the most important factors. Although we didn’t find that ROE was one of the most important factors, but the rule set generated by the regression trees implied that the higher the ROE, the higher the return. (4) The empirical results showed that the rules with the highest and lowest predicted return generated by regression trees achieved the highest and lowest return among all the rules, which showed that the regression tree can produce useful stock-picking rules.

碩士<br>輔仁大學<br>應用統計學研究所<br>95<br>Because the implementation of Basel II and large default loss of credit card and cash card effect in Taiwan, banks in Taiwan emphasized on the risk of various kinds of financial commodities. The topic of this research will discuss that the risk modeling method for car loan, and try to find the important factors. In this research, data for modeling were collected from a commercial bank in Taiwan during 2003 to 2005. The data includes 1,247 default customers of car loan in totally 17,292 car loan customers. In order to build better models, this research tried to use Weight of Evidence (WOE) scoring and ordinal scoring to transform categories variables into continuous variables. After transforming, logistic regression was used. At last, the research will compare with these models. The compared criterions are fitting statistics and prediction abilities of model including accuracy Rate, True Negative Rate and Recall Rate. After 5 times sampling comparing, the untransformed model is better than others. Finally, this research compared with four models. There are untransformed model, WOE scoring transformed model, ordinal scoring transformed model, and mixed transformed variables model with revised stepwise logistic regression. The mixed transformed variables model with revised stepwise logistic regression was the better model and was determined to be the best model by the definition of this research. In the best model, Accuracy Rate is 74.44%,True Negative Rate is 74.42%，and Recall Rate is 74.74%. After 15 times modeling, it can be found important and stable variables including to “method of loan”, “place of confirm”, “education levels”, “pay status”, “Number of other banks”, “inquire of other banks last three months”, ”the rate of short term repaid”, “balance of credit”, and “the rate of long term prepaid”.

This study investigates the association between traditional and alternative working capital measures and the returns of industrial firms listed on the Johannesburg Stock E"change. Twenty five variables for all industrial firms listed for the most recent 10 years were derived from standardised annual balance sheet data of the University of Pretoria's Bureau of Financial Analysis. Traditional liquidity ratios measuring working capital position, activity and leverage, and alternative liquidity measures, were calculated for each of the 135 participating firms for the 1 0 years. These working capital measures were tested for association with five return measures for every firm over the same period. This was done by means of a chi-square test for association, followed by stepwise multiple regression undertaken to quantify the underlying structural relationships between the return measures and the working capital measures. The results of the tests indicated that the traditional working capital leverage measures, in particular, total current liabilities divided by funds flow, and to a lesser e"tent, long-term loan capital divided by net working capital, displayed the greatest associations, and e"plained the majority of the variance in the return measures. At-test, undertaken to analyse the size effect on the working capital measures employed by the participating firms, compared firms according to total assets. The results revealed significant differences between the means of the top quartile of firms and the bottom quartile, for eight of the 13 working capital measures included in the study. A nonparametric test was applied to evaluate the sector effect on the working capital measures employed by the participating firms. The rank scores indicated significant differences in the means across the sectors for si" of the 13 working capital measures. A decrease in the working capital leverage measures of current liabilities divided by funds flow, and long-term loan capital divided by net working capital, should signal an increase in returns, and vice versa. It is recommended that financial managers consider these findings when forecasting firm returns.<br>Business Management<br>D. Com. (Business Management)

Trabalho de projeto de mestrado em Matemática Aplicada à Economia e à Gestão, apresentada à Universidade de Lisboa, através da Faculdade de Ciências, 2013<br>Este trabalho consiste num relatório de estágio, para obtenção do grau de mestre em Matemática Aplicada à Economia e Gestão, pela Faculdade de Ciências da Universidade de Lisboa. Estágio este realizado no Centro de Estudos e Avaliação em Saúde (CEFAR) sob a orientação de Zilda Mendes. No presente trabalho será apresentado um estudo de mercado realizado pelo CEFAR, onde o principal objectivo é avaliar a posição da marca em relação à concorrência no mercado de venda de produtos para o tratamento de Eczema/Eczema atópico e Psoríase. Será aplicado um modelo de regressão logística onde a variável resposta é o consumo, ou não, de produtos de uma determinada marca, identificando o conjunto de variáveis que melhor explicam a utilização desses produtos. Este trabalho é composto por 4 capítulos distintos. No primeiro capítulo, faremos uma breve introdução ao estudo de mercado, ilustrando o objectivo e as áreas de aplicação desse estudo. Seguidamente, apresentaremos o CEFAR, expondo a metodologia que utilizaram para a recolha de informação, a dimensão da amostra e as variáveis em estudo. No capítulo seguinte, introduziremos a análise de regressão logística univariada e posteriormente a regressão logística multivariada. Iniciaremos, em seguida, a análise prática, fazendo a análise descritiva dos dados, comparando os utilizadores dos produtos da concorrência com os utilizadores dos produtos da marca. Após esta análise, aplicar-se-á um modelo de regressão logística utilizando o método stepwise para a seleção das covariáveis de interesse. Por fim, serão apresentadas as conclusões do estudo.<br>This internship report was developed in order to obtain the Master Degree in “Matemática Aplicada à Economia e à Gestão”, at Faculdade de Ciências, Universidade de Lisboa. This internship was carried out in Centre for Health Evaluation & Research (CEFAR), under the supervision of MSc Zilda Mendes. Throughout this report it will be described a market research performed by CEFAR, where the main purpose was to assess the market position, of a determined brand, towards competitors in products for the treatment of Eczema/ Atopic Eczema and Psoriasis. This report focuses on the utilization of Logistic Regression Methods where the outcome variable is binary, consumption or not of products of a determined brand, identifying the best fitting model to describe the relationship between the outcome variable and the set of independent variable. This report is organized into four different chapters. In the first one, a brief introduction of market research, showing the purpose and application areas of these studies. Afterwards CEFAR was introduced, showing methods used for data collection, sample size estimation and definition of variables under study. In the second chapter was introduced logistic regression model in the univariate context and then for the multivariate case. Through the third chapter the practical performance was described, making a descriptive analysis of the data, comparing competitive products users with the brand users. After this analysis, model of logistic regression was applied using the stepwise method for selection of covariates of interest. Finally, in the fourth and last chapter the study final conclusions were reported.

Trabalho de Projeto do Mestrado Integrado em Engenharia Biomédica apresentado à Faculdade de Ciências e Tecnologia<br>A produção de scaffolds para engenharia de tecidos é uma área de investigação, cujo interesse tem aumentado ao longo do tempo. A procura de métodos ''verdes'' que promovam tanto processos eficientes a nível energético, como a redução de desperdícios, tem requerido metodologias ótimas e que garantam qualidade. Tomando como inspiração o exemplo de outras indústrias de produção, como a indústria farmacêutica que tem promovido a adoção de abordagens de qualidade pelo design para assim melhorar os seus processos, otimizando-os, este trabalho enaltece a importância de adotar uma abordagem semelhante para o processo de produção de um aerogel para engenharia de tecidos óssea. Os dados que foram recolhidos para este processo não foram obtidos conforme uma metodologia ótima, como é o caso do design de experiências e, como tal, foi mais difícil extrair a máxima informação a partir de um número limitado de experiências. Contudo, o objetivo principal foi o de encontrar uma forma de estimar os valores dos parâmetros capazes de otimizar o processo. Para isso, em primeiro lugar propôs-se uma divisão sistemática e rigorosa do processo. Assim, aplicou-se um método de seleção de variáveis através de uma regressão passo-a-passo com seleção em frente que levou à obtenção de modelos capazes de representar o processo em estudo. Posteriormente, aplicou-se uma otimização multi-objetivo, com recurso a funções desejáveis. A abordagem mencionada foi capaz de otimizar várias respostas em simultâneo, sendo estas propriedades específicas do aerogel. Consequentemente, com esta abordagem foi possível não só perceber como os parâmetros do processo influenciam o produto final, mas também como estes podem ser geridos para produzir um aerogel adequado a uma aplicação médica específica. Para além disto, é também possível monitorizar a produção tendo em conta as limitações do processo e/ou dos recursos.<br>The production of scaffolds for tissue engineering is a research area that has received increasing interest through the course of time. The search for greener manufacturing methods to promote energy-efficient processes, as well as waste reduction, has demanded for both optimal and quality-assuring methodologies. Inspired by the example of other manufacturing industries, such as the pharmaceutical industry, which has promoted the adoption of quality by design approaches to its processes in order to optimize them, this work highlights the importance of adopting such an approach to the specific process of aerogel production for bone tissue engineering. The data collected for this process was not obtained according to an optimal methodology such as the design of experiments, making it more difficult to extract the maximum amount of information from a limited number of experimental runs. Nonetheless, the main goal was to find a way of estimating the best settings of the parameters that optimized this process. Therefore, a rigorous and systematic division of the process was initially proposed. On that account, a variable selection procedure using stepwise regression with forward selection was applied, leading to models able to mimic the process studied. Afterwards, a multi-objective optimization using desirability functions was employed. This approach was able to simultaneously optimize several responses which were aerogel specific properties. Consequently, it was possible, not only to understand how the process parameters affect the final product, but also how they can be managed to produce aerogels fit for a specific medical application. Moreover, it can also monitor the production with regard to the process and/or resources limitations.<br>Outro - POCI-01-0145-FEDER-032625

This thesis aims to identify main driving market risk factors of different strategies implemented by hedge funds by looking at correlation coefficients, implementing Principal Component Analysis and analyzing "loadings" for first three principal components, which explain the largest portion of the variation of hedge funds' returns. In the next step, a stepwise regression through iteration process includes and excludes market risk factors for each strategy, searching for the combination of risk factors which will offer a model with the best "fit", based on The Akaike Information Criterion - AIC and Bayesian Information Criterion - BIC. Lastly, to avoid counterfeit results and overcome model uncertainty issues a Bayesian Model Average - BMA approach was taken. Key words: Hedge Funds, hedge funds' strategies, market risk, principal component analysis, stepwise regression, Akaike Information Criterion, Bayesian Information Criterion, Bayesian Model Averaging Author's e-mail: aleksaradosavcevic@gmail.com Supervisor's e-mail: mp.princ@seznam.cz

Trabalho de projeto de mestrado, Matemática aplicada à Economia e Gestão, Universidade de Lisboa, Faculdade de Ciências, 2017<br>Através de um estágio no Banco B na área de Data Mining e Estudos de Mercado, este trabalho foi feito para a obtenção do grau de mestre em Matemática Aplicada à Economia e Gestão, pela Faculdade de Ciências da Universidade de Lisboa. O presente trabalho consiste na realização de um Modelo de Regressão Logística com a finalidade de estudar a vinculação de um Cliente a um banco. Foi feito um estudo de mercado onde o principal objetivo é saber o perfil de cada Cliente que considera um determinado banco como o seu banco principal. O Modelo de Regressão Logística terá como variável resposta a vinculação, ou não, de um Cliente a um banco consoante as variáveis que melhor caracterizam os seus perfis. No Modelo Logístico existe a particularidade de, através de um conjunto de variáveis independentes, se conseguir prever a vinculação de um determinado Cliente. Esta é a melhor metodologia adotada para este estudo uma vez que interpreta o impacto marginal de cada variável na vinculação do Cliente. Como tal este trabalho é composto por 4 capítulos. No primeiro é feita a introdução aos conteúdos teóricos da regressão logística múltipla. No capítulo seguinte será explicado o objetivo deste estudo e uma análise descritiva dos dados. Em seguida será posto em prática o Modelo de Regressão Logística através do uso de dados reais e de técnicas como a discretização de variáveis, árvores de decisão, R-Quadrado, Qui-Quadrado e o Stepwise (método que seleciona as variáveis finais do modelo). Neste último capítulo é feita ainda uma avaliação do modelo e consequente análise do perfil de Clientes e validação do modelo construído. Por fim será apresentado o capítulo que conclui e comenta o trabalho feito ao longo do modelo e ainda sugere algumas alternativas metodológicas tais como Splines Cúbicas e ainda um modelo alternativo.<br>Through an internship at Bank B in the area of Data Mining and Market Research, this work was done to obtain a master’s degree in Mathematics Applied to Economics and Management, by the Faculty of Sciences of the University of Lisbon. The present work focuses on the utilization of a Logistic Regression Model with the purpose of studying the connection of a Client to a bank. A market research was done where the main objective is to know the profile of each Client that considers a bank as his/her main bank. The Logistic Regression Model will have as response variable the binding, or not, of a Client to a bank according to the variables that best characterize his/her profile. In the Logistic Model, there is the particularity of, through a set of independent variables, predicting the binding of a certain Client. This is the best methodology adopted for this study since it interprets the marginal impact of each variable in the Customer's binding. As such this work is composed of 4 chapters. In the first chapter, the theoretical contents of Multiple Logistic Regression are introduced. In the following chapter, we will explain the purpose of this study and a descriptive analysis of the data. Then, in the third chapter, the Logistic Regression Model will be implemented using real data and techniques such as discretization of variables, decision trees, R-Square, Chi-Square and Stepwise (method that selects the final variables of the model). In this last chapter is done an evaluation of the model and consequently analysis of the profile of each Client and validation of the built model. Finally, in the fourth, we will present the chapter that concludes and comments all the work done throughout the model and suggests some methodological alternatives such as Cubic Splines and an alternative model.

Die studie vergelyk die prestasie van 'n koop-en-hou-strategie met die van 'n tydsberekeningstrategie binne die raamwerk van die arbitrasie-prysbepalingsteorie (APT) op die nywerheidsindeks van die Johannesburgse Aandelebeurs (JA). Die periode van die studie is oor twee tydperke, naamlik Januarie 1970 tot September 1987 en Januarie 1989 tot Junie 1997. Die langtermyntendens van die nywerheidsindeks en APT-faktore is bepaal deur die beste nie-reglynige model vir elke tydreeks te vind. Reglynige meervoudige stapsgewyse regressie-ontleding is gebruik om die bewegings van die nywerheidsindeks rondom die langtermyntendens te voorspel. Die sloeringsreekse van die langtermyntendensresidutelling van die APT-faktore en die sloeringsreekse van die eerste-ordeverskiltelling van die langtermyntendensresidutelling is as moontlike voorspellers gebruik. Gegrond hierop is beslissingslyne ontwik:kel wat gebruik is vir die implementering van 'n tydsberekeningstrategie. Die resultate van die studie is die volgende: • Waar die sloeringsreekse van die langtermyntendensresidutelling van die APTfaktore as moontlike voorspellers gebruik is, is die risiko-aangepaste opbrengskoers van 'n tydsberekeningstrategie 6, 41 persent en 0, 71 persent b6 die van 'n koop-en-hou-strategie vir tydperk een en twee onderskeidelik. • Waar die sloeringsreekse van die eerste-ordeverskiltelling van die langtermyntendensresidutelling van die APT-faktore as moontlike voorspellers gebruik is, is die risiko-aangepaste opbrengskoers van 'n tydsberekeningstrategie 10,40 persent en 1,04 persent b6 die van 'n koop-enhou- strategie vir tydperk een en twee onderskeidelik. Die belangrikste gevolgtrekking van die studie is dat die APT en 'n tydsberekeningstrategie teoreties en prakties versoenbaar is op die JA. Aanbevelings vir toekomstige navorsing is die volgende: ( 1) sistematiese risikofaktore, anders as makro-ekonomiese faktore, behoort identifiseer te word wat die voorspellingswaarde van die faktore in die tweede tydperk van die studie kan verhoog; (2) elke stap van die model wat ontwikkel is, behoort op elke indeks van die JA toegepas te word om die risiko-aangepaste opbrengskoers van 'n tydsberekeningstrategie toegepas op elkeen van die indekse met die van 'n koop-en-hou-strategie te vergelyk; en (3) die invloed van transaksiekoste en dividende op die potensiele voordele van tydsberekening moet bepaal word.<br>The study compares the performance of a buy-and-hold strategy with that of a markettiming strategy in the framework of the arbitrage pricing theory (APT) applied to the industrial index of the Johannesburg Stock Exchange (JSE). The study period is divided into two parts, namely January 1970 to September 1987 and January 1989 to June 1997. The long-term trend of the industrial index and every APT factor is determined by finding the best nonlinear model for each time series. Linear multiple stepwise regression analysis, with the lagged time series of the long-term trend error terms of the APT factors, is used to forecast the movement of the industrial index around its long-term trend. Decision lines were developed to implement a market-timing strategy. The results of the study are as follows: • Where the lagged time series of the long-term trend error terms of the APT factors were used as possible predictors, the risk-adjusted return of a markettiming strategy was 6, 41 percent and 0, 71 percent higher than that of a buyand- hold strategy for periods one and two respectively. • Where the lagged time series of the first-order difference of the long-term trend error term of the APT factors were used as possible predictors, the riskadjusted return of the market-timing strategy was 10,40 percent and 1,04 percent higher than that of a buy-and-hold strategy for periods one and two respectively. The main conclusion of the study is that the APT and a market-timing strategy are theoretically and practically reconcilable on the JSE. The main recommendations of the study are the following: (1) systematic risk factors, other than macroeconomic factors, should be identified in order to increase the forecasting value of these factors in the second period of the study; (2) each step of the model developed in this study should be repeated on every index of the JSE; and (3) the influence of transaction costs and dividends on the potential benefits of a market-timing strategy should be determined.<br>Business Management<br>DCom (Sakebestuur)