第一章 緒論 第一節 研究背景 第二節 研究問題 第三節 研究目的 第四節 名詞定義及研究範圍 第二章 評估信用風險之文獻參考 第三章 研究設計 第一節 研究架構與變數 第二節 抽樣設計 第三節 資料分析方法 第四章 資料分析與研究結果 第一節 企業主持人執業資格、經驗、地緣關係與信用風險之關係。 第二節 各項特質對信用風險之預測能力。 第五章 結論與建議 附錄 Log-Iknear Models 統計技術簡介。

The purpose of this paper is to generalize regression models for repeated categorical data based on maximizing a conditional likelihood. Some existing methods, such as those proposed by Duncan (1985), Fischer (1989), and Agresti (1993, and 1997) are special cases of this latent variable approach, used to account for dependencies in clustered observations. The generalization concerns the incorporation of rather general data structures such as subject-specific time-dependent covariates, a variable number of observations per subject and time periods of arbitrary length in order to evaluate treatment effects on a categorical response variable via a linear parameterization. The response may be polytomous, ordinal or dichotomous. The main tool is the log-linear representation of appropriately parameterized Rasch-type models, which can be fitted using standard software, e.g., R. The proposed method is applied to data from a psychiatric study on the evaluation of psychobiological variables in the therapy of depression. The effects of plasma levels of the antidepressant drug Clomipramine and neuroendocrinological variables on the presence or absence of anxiety symptoms in 45 female patients are analyzed. The individual measurements of the time dependent variables were recorded on 2 to 11 occasions. The findings show that certain combinations of the variables investigated are favorable for the treatment outcome. (author´s abstract) / Series: Research Report Series / Department of Statistics and Mathematics

Reliability issues are always salient as behavioral researchers observe human behavior and classify individuals from criterion-referenced test scores. This has created a need for studies to assess agreement between observers, recording the occurrance of various behaviors, to establish the reliability of their classifications. In addition, there is a need for measuring the consistency of dichotomous and polytomous classifications established from criterion-referenced test scores. The development of several log linear univariate models for measuring agreement has partially met the demand for a probability-based measure of agreement with a directly interpretable meaning. However, multi-variate repeated measures agreement produres are necessary because of the development of complex intrasubject and intersubject research designs. The present investigation developed applications of the log linear, latent class, and weighted least squares procedures for the analysis of multivariate repeated measures designs. These computations tested the model-data fit and calculated the multivariate measure of the magnitude of agreement under the quasi-equiprobability and quasi-independence models. Applications of these computations were illustrated with real and hypothetical observational data. It was demonstrated that employing log linear, latent class, and weighted least squares computations resulted in identical multi-variate model-data fits with equivalent chi-square values. Moreover, the application of these three methodologies also produced identical measures of the degree of agreement at each point in time and for the multivariate average. The multivariate methods that were developed also included procedures for measuring the probability of agreement for a single response classification or subset of classifications from a larger set. In addition, procedures were developed to analyze occurrences of systematic observed disagreement within the multivariate tables. The consistency of dichotomous and polytomous classifications over repeated assessments of the identical examinees was also suggested as a means of conceptualizing criterion-referenced reliability. By applying the univariate and multivariate models described, the reliability of these classifications across repeated testings could be calculated. The procedures utilizing the log linear, latent structure, and weighted least squares concepts for the purpose of measuring agreement have the advantages of (1)yielding a coefficient of agreement that varies between zero and one and measures agreement in terms of the probability that the observers' judgements will agree, as estimated under a quasi-equiprobability or quasi-independence model, (2)correcting for the proportion of "chance" agreement, and (3) providing a directly interpretable coefficient of "no agreement." Thus, these multivariate procedures may be regarded as a more refined psychometric technology for measuring inter-observer agreement and criterion-referenced test reliability.

The purpose of this study was to investigate the performance of the parametric bootstrap method and to compare the parametric and nonparametric bootstrap methods for estimating the standard error of equating (SEE) under the common-item nonequivalent groups (CINEG) design with the frequency estimation (FE) equipercentile method under a variety of simulated conditions. When the performance of the parametric bootstrap method was investigated, bivariate polynomial log-linear models were employed to fit the data. With the consideration of the different polynomial degrees and two different numbers of cross-product moments, a total of eight parametric bootstrap models were examined. Two real datasets were used as the basis to define the population distributions and the "true" SEEs. A simulation study was conducted reflecting three levels for group proficiency differences, three levels of sample sizes, two test lengths and two ratios of the number of common items and the total number of items. Bias of the SEE, standard errors of the SEE, root mean square errors of the SEE, and their corresponding weighted indices were calculated and used to evaluate and compare the simulation results. The main findings from this simulation study were as follows: (1) The parametric bootstrap models with larger polynomial degrees generally produced smaller bias but larger standard errors than those with lower polynomial degrees. (2) The parametric bootstrap models with a higher order cross product moment (CPM) of two generally yielded more accurate estimates of the SEE than the corresponding models with the CPM of one. (3) The nonparametric bootstrap method generally produced less accurate estimates of the SEE than the parametric bootstrap method. However, as the sample size increased, the differences between the two bootstrap methods became smaller. When the sample size was equal to or larger than 3,000, the differences between the nonparametric bootstrap method and the parametric bootstrap model that produced the smallest RMSE were very small. (4) Of all the models considered in this study, parametric bootstrap models with the polynomial degree of four performed better under most simulation conditions. (5) Aside from method effects, sample size and test length had the most impact on estimating the SEE. Group proficiency differences and the ratio of the number of common items to the total number of items had little effect on a short test, but had slight effect on a long test.

Abstract Pedestrian /bicycle safety of school children has been a growing menace that has been attracting attention from transportation professionals, school boards, media and the community all over the country. As such there has been a necessity to identify critical variables and assess their importance in pedestrian/bicycle crashes occurring in and around school zones. The current study is an endeavor in this direction. The literature review identified some studies that were conducted on school zone safety related to pedestrian/bicyclist crashes. Most of the studies pertained to crashes with all age groups. There have been few studies with emphasis only on school aged children. In this study we focus on pedestrian age group (4 to 18 years), the time of the day when the school children are expected to be commuting (6:30 AM to 10:00 AM and 1:00 PM to 5:00.PM), the day of week (Monday through Friday) and the days when the school is opened (January 6th to May 31st and August 6th to December 21st). Geographical Information Systems was used to locate buffer zones around schools with higher crash incidence rates. The use of log-linear analysis has culminated in explaining the relationship between various variables and crash incidence or crash frequency Crash data for this study was obtained in the form of crash database and GIS maps from the Department of Highway Safety and Motor Vehicles and the Orange County School Board respectively. Crash reports were downloaded from the CAR database of the FDOT mainframe website. The crash data was related to the GIS maps to visually depict the proximity of crashes to the school zones and thus identified risky schools and school districts. It was concluded from the spatial analysis that the incidence of crashes was higher at middle schools. In the log-linear analysis different models were i tested to explain the effects of driver characteristics, geometric characteristics and pedestrian characteristics on the crash frequency. It was found that driver age, number of lanes, median type, pedestrian age and speed limit are the critical variables in explaining crash frequency. By examining the levels of the variables that were significantly involved in the crashes we would get an insight on ways to explain and control pedestrian/bicyclists crashes at school zones. It is hoped that this thesis would make an active contribution in improving the safety of bicyclists and pedestrians in and around school zones and make the schools much safer for the children.

多項式回歸模型可用來模擬賽馬過程。不同研究者對模型中馬匹的效用的分佈採取不同的假設，包括指數分佈，它與Harville 模型（Harville, 1973）相同，伽馬分佈（Stern, 1990）和正態分佈（Henery, 1981）。Harville 模型無法模擬賽馬過程中競爭第二位和第三位等非冠軍位置時增加的隨機性（Benter, 1994）。Stern 模型假設效用服從形狀參數大於一的伽馬分佈，Henery 模型假設效用服從正態分佈。Bacon-Shone，Lo 和 Busche（1992），Lo 和 Bacon-Shone（1994）和 Lo（1994）研究證明了相較於Harville 模型，這兩個模型能更好地模擬賽馬過程。本文利用賽馬歷史數據，採用貝葉斯方法對賽馬結果中馬匹勝出的概率進行預測。本文假設效用服從伽馬分佈。本文針對多項式回歸模型，提出一個在Metropolis-Hastings 抽樣方法中選擇提議分佈的簡便方法。此方法由Scott（2008）首次提出。我們在似然函數中加入服從伽馬分佈的效用作為潛變量。通過將服從伽馬分佈的效用變換成一個服從Mihram（1975）所描述的廣義極值分佈的隨機變量，我們得到一個線性回歸模型。由此線性模型我們可得到最小二乘估計，本文亦討論最小二乘估計的漸進抽樣分佈。我們利用此估計的方差得到Metropolis-Hastings 抽樣方法中的提議分佈。最後，我們可以得到回歸參數的後驗分佈樣本。本文用香港賽馬數據做模擬賽馬投資以檢驗本文提出的估計方法。 / In multinomial regression of racetrack betting, dierent distributions of utilities have been proposed: exponential distribution which is equivalent to Harville’s model (Harville, 1973), gamma distribution (Stern, 1990) and normal distribution (Henery, 1981). Harville’s model has the drawback that it ignores the increasing randomness of the competitions for the second and third place (Benter, 1994). The Stern’s model using gamma utilities with shape parameter greater than 1 and the Henery’s model using normal utilities have been shown to produce a better t (Bacon-Shone, Lo and Busche, 1992; Lo and Bacon-Shone, 1994; Lo, 1994). In this thesis, we use the Bayesian methodology to provide prediction on the winning probabilities of horses with the historical observed data. The gamma utility is adopted throughout the thesis. In this thesis, a convenient method of selecting Metropolis-Hastings proposal distributions for multinomial models is developed. A similar method is rst exploited by Scott (2008). We augment the gamma distributed utilities in the likelihood as latent variables. The gamma utility is transformed to a variable that follows generalized extreme value distribution described by Mihram (1975) through which we get a linear regression model. Least squares estimate of the parameters is easily obtained from this linear model. The asymptotic sampling distribution of the least squares estimate is discussed. The Metropolis-Hastings proposal distribution is generated conditioning on the variance of the estimator. Finally, samples from the posterior distribution of regression parameters are obtained. The proposed method is tested through betting simulations using data from Hong Kong horse racing market. / Detailed summary in vernacular field only. / Xu, Wenjun. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2012. / Includes bibliographical references (leaves 46-48). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstracts also in Chinese. / Chapter 1 --- Introduction --- p.1 / Chapter 2 --- Hong Kong Horse Racing Market and Models in Horse Racing --- p.4 / Chapter 2.1 --- Hong Kong Horse Racing Market --- p.4 / Chapter 2.2 --- Models in Horse Racing --- p.6 / Chapter 3 --- Metropolis-Hastings Algorithm in Multinomial Regression with Gamma Utilities --- p.10 / Chapter 3.1 --- Notations and Posterior Distribution --- p.10 / Chapter 3.2 --- Metropolis-Hastings Algorithm --- p.11 / Chapter 4 --- Application --- p.15 / Chapter 4.1 --- Variables --- p.16 / Chapter 4.2 --- Markov Chain Simulation --- p.17 / Chapter 4.3 --- Model Selection --- p.27 / Chapter 4.4 --- Estimation Result --- p.31 / Chapter 4.5 --- Betting Strategies and Comparisons --- p.33 / Chapter 5 --- Conclusion --- p.41 / Appendix A --- p.43 / Appendix B --- p.44 / Bibliography --- p.46

Thesis (M.Sc. (Statistics)) - Rhodes University, 2007. / A thesis submitted in partial fulfilment of the requirements for the degree of Master of Science.

The objective of this project is to link the statistical theory used in the ecological sciences with an actual project that was developed for the South African National Parks Scientific Services. It investigates the changes that have occurred in the herbaceous and woody structure due to the closure of artificial waterpoints; including the impacts that elephants and other herbivores have on the vegetation of the Kruger National Park. This project was designed in conjunction with South African National Parks (SANP) Scientific Services and it is a registered project with this department. The results of this project will be submitted to Scientific Services in accordance with the terms and conditions of a SANP research project. A major concern within the KNP is the declining numbers of rare antelope and numerous projects have been developed to investigate possible ways of halting this decline and thus protecting the heterogeneity of the Kruger National Park. Three different datasets were investigated, covering three aspects of vegetation structure and composition within the KNP. The first investigated the changes that have occurred since the N'washitsumbe enclosure in the Far Northern KNP was fenced off from the rest of the park. The results show that over the 40 years since the enclosure was built, changes have occurred which have resulted in a significant difference in the abundance of Increaser 2 and Decreaser grass species between the inside and the outside of the enclosure. Increaser 2 and Decreaser categories are the result of a grass species classification depending on whether the species thrives or is depressed by heavy grazing. The difference in grass species composition and structure between the inside and the outside of the enclosure indicates that the grazing animals within the KNP have influenced the grass composition in a way that favours the dominant animals. This has resulted in a declining roan antelope population - one of the species that is considered as a 'rare antelope'. Many artificial waterpoints (boreholes and dams) have also been closed throughout the KNP in the hope of resulting in a change in vegetation structure and composition in favour of the roan. Veld condition assessment data for 87 boreholes throughout the Park was analyzed to determine whether the veld in the vicinity is beginning to change towards a more Decreaser dominated sward which would favour the roan. The results were analyzed for the different regions of the Park; and they indicate that changes are becoming evident; however, the results are not particularly conclusive, yet. The majority of the boreholes were closed between 1994 and 1998 which means that not a lot of data were available to be analyzed. A similar study conducted in another 10 years time might reveal more meaningful results. However the results are moving in the direction hoped for by the management of the KNP. The results show that the grass composition has a higher proportion of Decreaser grasses since the closure of the waterpoints, and the grass biomass around these areas has also improved. The results were analyzed on an individual basis; and then on a regional basis as the minimal data meant that the individual analyses did not provide any significant results. A third study was then done on the impact that the rapidly increasing elephant population on the vegetation within the Riparian zone along three rivers in the Far Northern region of the KNP. The riparian zone is an important part of the landscape, in terms of providing food for many animals as well as shade. The elephant population has increased substantially since the termination of the culling program and this means that the feeding requirements of the population has increased which could result in severe damage upon the vegetation, as elephants can be very destructive feeders. The results show surprising differences between the three years of data that were analyzed; however the results indicate that the elephants are targeting specific height ranges of trees when feeding; however they do not seem to consistently target specific tree species. This is positive for the diversity of the Riparian zone as this region is very important both ecologically and aesthetically for the tourists who visit the Park.

A sample of corporate instructional designers and professors of instructional design completed the "Corporate Instructional Design Scale." The data yielded information on the extent of agreement that descriptive statements identified conventionally and systematically designed instruction. Descriptive and asymmetric log linear (statistical) analyses were conducted. In the asymmetric log linear analyses, the extent of agreement was used as the dependent variable. The three independent variables with three levels each were Program type (conventionally designed instruction, both conventionally and systematically designed instruction, and systematically designed instruction), Instructional component (instructional intents, instructional strategies, and instructional assessments), and Trainer type (professional trainers in manufacturing, professional trainers in non-manufacturing, and professors of instructional design). The asymmetric log linear analysis using 16 models was a 3x3x3x3 factorial design. The extent of agreement on the indicators of conventional instruction was lower than the extent of agreement on the indicators of systematic instruction. The extent of agreement for instructional assessment indicators was lower than the extent of agreement for instructional intents and strategies. There were only minor differences between the extent of agreement on indicators classified as intents and indicators classified as strategies. the extent of agreement on the indicators which differentiated conventionally and systematically designed instruction was higher for the professors of instructional design than for the trainers in manufacturing and non-manufacturing companies. Study results should be carefully considered by professors of instructional design when designing their instructional design courses. The high extent of agreement by professors of instructional design on items that distinguished conventional instruction and systematic instruction suggest that academia is fairly clear about the indicators of instructional design, specially instructional intents and instructional strategies, while the practitioners of instructional design have a substantially lower extent of agreement. These results suggest at least two conclusions. First, the academic world of instructional design is not in tune with the corporate world. Academia has been promoting idealized procedures for instructional design, while practitioners have adjusted their instructional designed to corporate realities of time and cost. Second, corporate instructional designers have found academic world suggestions unrealistic. Corporate instructional designers have made modifications to their instructional designs. Their instructional designs may actually only approximate whatever type of instruction the professional trainers or corporation where they are employed may advocate.

Esta tesis reune algunas contribuciones al reconocimiento de formas estadístico y, más especícamente, a varias tareas del procesamiento del lenguaje natural. Varias técnicas estadísticas bien conocidas se revisan en esta tesis, a saber: estimación paramétrica, diseño de la función de pérdida y modelado estadístico. Estas técnicas se aplican a varias tareas del procesamiento del lenguajes natural tales como clasicación de documentos, modelado del lenguaje natural y traducción automática estadística. En relación con la estimación paramétrica, abordamos el problema del suavizado proponiendo una nueva técnica de estimación por máxima verosimilitud con dominio restringido (CDMLEa ). La técnica CDMLE evita la necesidad de la etapa de suavizado que propicia la pérdida de las propiedades del estimador máximo verosímil. Esta técnica se aplica a clasicación de documentos mediante el clasificador Naive Bayes. Más tarde, la técnica CDMLE se extiende a la estimación por máxima verosimilitud por leaving-one-out aplicandola al suavizado de modelos de lenguaje. Los resultados obtenidos en varias tareas de modelado del lenguaje natural, muestran una mejora en términos de perplejidad. En a la función de pérdida, se estudia cuidadosamente el diseño de funciones de pérdida diferentes a la 0-1. El estudio se centra en aquellas funciones de pérdida que reteniendo una complejidad de decodificación similar a la función 0-1, proporcionan una mayor flexibilidad. Analizamos y presentamos varias funciones de pérdida en varias tareas de traducción automática y con varios modelos de traducción. También, analizamos algunas reglas de traducción que destacan por causas prácticas tales como la regla de traducción directa; y, así mismo, profundizamos en la comprensión de los modelos log-lineares, que son de hecho, casos particulares de funciones de pérdida. Finalmente, se proponen varios modelos de traducción monótonos basados en técnicas de modelado estadístico . / Andrés Ferrer, J. (2010). Statistical approaches for natural language modelling and monotone statistical machine translation [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/7109 / Palancia