This dissertation used synthetic datasets, semi-synthetic datasets, and a real-world dataset from an educational intervention to compare the performance of 15 machine learning and multiple imputation methods to estimate the individual treatment effect (ITE). In addition, it examined the performance of five evaluation metrics that can be used to identify the best ITE estimation method when conducting research with real-world data. Among the ITE estimation methods that were analyzed, the S-learner, the Bayesian Causal Forest (BCF), the Causal Forest, and the X-learner exhibited the best performance. In general, the meta-learners with BART and tree-based direct estimation methods performed better than the representation learning methods and the multiple imputation methods. As for the evaluation metrics, τ_(risk_R ) and the Switch Doubly Robust MSE (SDR-MSE) performed the best in identifying the best ITE estimation method when the true treatment effect was unknown. This dissertation contributes to a small but growing body of research on ITE estimation which is gaining popularity in various fields due to its potential for tailoring interventions to meet the needs of individuals and targeting programs at those who would benefit the most from those interventions.

京都大学 / 新制・論文博士 / 博士(社会健康医学) / 乙第13612号 / 論社医博第18号 / 新制||社医||13(附属図書館) / 京都大学大学院医学研究科社会健康医学系専攻 / (主査)教授 森田 智視, 教授 古川 壽亮, 教授 今中 雄一 / 学位規則第4条第2項該当 / Doctor of Public Health / Kyoto University / DFAM

Water temperature is a critical factor for the quality and biological condition of streams. Among various factors affecting stream water temperature, air temperature is one of the most important factors related to water temperature. To appropriately quantify the relationship between water and air temperatures over a large geographic region, it is important to accommodate the spatial and temporal information of the steam temperature. In this dissertation, I devote effort to several statistical modeling techniques for analyzing bivariate spatial-temporal data in a stream temperature study. In the first part, I focus our analysis on the individual stream. A time varying coefficient model (VCM) is used to study the relationship between air temperature and water temperature for each stream. The time varying coefficient model enables dynamic modeling of the relationship, and therefore can be used to enhance the understanding of water and air temperature relationships. The proposed model is applied to 10 streams in Maryland, West Virginia, Virginia, North Carolina and Georgia using daily maximum temperatures. The VCM approach increases the prediction accuracy by more than 50% compared to the simple linear regression model and the nonlinear logistic model. The VCM that describes the relationship between water and air temperatures for each stream is represented by slope and intercept curves from the fitted model. In the second part, I consider water and air temperatures for different streams that are spatial correlated. I focus on clustering multiple streams by using intercept and slope curves estimated from the VCM. Spatial information is incorporated to make clustering results geographically meaningful. I further propose a weighted distance as a dissimilarity measure for streams, which provides a flexible framework to interpret the clustering results under different weights. Real data analysis shows that streams in same cluster share similar geographic features such as solar radiation, percent forest and elevation. In the third part, I develop a spatial-temporal VCM (STVCM) to deal with missing data. The STVCM takes both spatial and temporal variation of water temperature into account. I develop a novel estimation method that emphasizes the time effect and treats the space effect as a varying coefficient for the time effect. A simulation study shows that the performance of the STVCM on missing data imputation is better than several existing methods such as the neural network and the Gaussian process. The STVCM is also applied to all 156 streams in this study to obtain a complete data record. / Ph. D.

Implementing Lean in the United States transportation equipment manufacturing industry holds the promise for improvements in, among other things, productivity, quality, and innovation, resulting in more competitive success and profits. Although Lean has been applied throughout the industry with noted success, there have been some difficulties in demonstrating the financial benefits derived from Lean initiatives. Most of the evidence supporting a positive relationship between Lean implementation and improved financial performance is anecdotal. As companies have become more proficient in carrying out Lean initiatives in manufacturing, they have extended Lean ideas to other parts of their organization and throughout the entire supply chain. Nowadays, it is widely recognized that a holistic, enterprise-wide view is critical to obtain the potential benefits of a Lean transformation. However, Lean transformations are often undertaken without consideration of supporting functions such as accounting and finance. Lean transformation in accounting and finance should be run in the same way as it is in the manufacturing environment by decreasing reporting cycle time, improving transaction processing accuracy, eliminating unnecessary transaction processing, changing product costing procedures, and financial reporting among many other things, but there is limited empirical evidence of that happening. To address these shortcomings, this research focuses on three areas. First, this study aims to evaluate transportation equipment manufacturing facilities in respect to their operational and financial performance. Second, this study aims to investigate the extent of Lean implementation of a given operation in respect to leadership, manufacturing, accounting and finance, and supplier and customer relationship and correlate these results to their performance. Finally, this study aims to further examine the contextual characteristics of companies that successfully aligned their systems with Lean. A mixed-mode survey, addressed to a subset of the United States transportation equipment manufacturing industry, asked questions pertinent to companies' Lean transformation efforts, performance, and general characteristics. During the four months long survey period, a total of 69 valid responses were received, for a response rate of 3.78 percent. From the 69 valid responses, 8 responses were eliminated due to containing more than 20 percent missing values. Multiple imputation procedure was applied to handle remaining missing values in the dataset. Before testing study hypotheses, scale reliability and construct validity tests were run to decide whether a particular survey item should be retained in further analysis. Study hypotheses were then tested using profile deviation analysis, multiple regression analysis, and hierarchical regression analysis. When the level of Lean implementation and performance relationship was investigated using a multiple regression analysis, results did not show any evidence that the higher level of Lean implementation along four business dimensions (leadership, manufacturing, accounting and finance, and supplier and customer relationship) of transportation equipment manufacturing facilities positively influences their operational and financial performance. However, it was revealed that the higher level of Lean implementation in transportation equipment manufacturing facilities' manufacturing dimension resulted in better quality performance as measured by first-time through, inbound quality, and outbound quality. When the same relationship was investigated using a profile deviation analysis, results were identical. When the level of Lean implementation in accounting and finance and its relationship with performance was investigated using a single regression analysis, results showed that the higher level of Lean implementation in transportation equipment manufacturing facilities' accounting and finance dimension has a positive effect on accounting performance and on operational performance (e.g., on time-based performance and delivery-based performance), but no effect on financial performance. When the same relationship was investigated using a profile deviation analysis, results were different by showing no relationship between the level of Lean implementation in transportation equipment manufacturing facilities' accounting and finance dimension and accounting, operational, and financial performance. Lastly, the effect of contextual variables (e.g., industry segment, location, annual sales volume, and unionization) on performance, the level of Lean implementation, and the performance -- Lean implementation relationship was investigated using hierarchical regression. Results showed that transportation equipment manufacturing facilities' performance is influenced by annual sales volume. Their level of Lean implementation in the accounting and finance dimension is influenced by location, while their performance -- Lean implementation in the accounting and finance dimension relationship is influenced by industry segment. / Ph. D.

Genetic association studies usually include related individuals. Meanwhile, high-throughput sequencing technologies produce data of multiple genetic variants. Due to linkage disequilibrium (LD) and familial relatedness, the genotype data from such studies often carries complex correlations. Moreover, missing values in genotype usually lead to loss of power in genetic association tests. Also, repeated measurements of phenotype and dynamic covariates from longitudinal studies bring in more opportunities but also challenges in the discovery of disease-related genetic factors. This dissertation focuses on developing novel statistical methods to address some challenging questions remaining in genetic association studies due to the aforementioned reasons. So far, a lot of methods have been proposed to detect disease-related genetic regions (e.g., genes, pathways). However, with multiple-variant data from a sample with relatedness, it is critical to account for the complex genotypic correlations when assessing genetic contribution. Recognizing the limitations of existing methods, in the first work of this dissertation, the Adaptive-weight Burden Test (ABT) --- a score test between a quantitative trait and the genotype data with complex correlations --- is proposed. ABT achieves higher power by adopting data-driven weights, which make good use of the LD and relatedness. Because the null distribution has been successfully derived, the computational simplicity of ABT makes it a good fit for genome-wide association studies. Genotype missingness commonly arises due to limitations in genotyping technologies. Imputation of the missing values in genotype usually improves quality of the data used in the subsequent association test and thus increases power. Complex correlations, though troublesome, provide the opportunity to proper handling of genotypic missingness. In the second part of this dissertation, a genotype imputation method is developed, which can impute the missingness in multiple genetic variants via the LD and the relatedness. The popularity of longitudinal studies in genetics and genomics calls for methods deliberately designed for repeated measurements. Therefore, a multiple-variant genetic association test for a longitudinal trait on samples with relatedness is developed, which treats the longitudinal measurements as observations of functions and thus takes into account the time factor properly. / PHD / It has been widely recognized that complex diseases are results of poor habits and genetic predisposition. Though people can make their own choices about lifestyle, the mysterious genome language seems to be unchangeable and inevitable. Decoding the messages delivered by DNA can help with prevention, prediction and treatment of diseases. This work focuses on developing novel statistical methods that can make contributions to the detection of disease-related genetic factors. Specifically, given the genotype data and phenotype (e.g., fasting glucose level) data on a sample of individuals where some could be relatives and the rest may be not, three challenges are addressed in this work: (1) how to detect if a genetic region (such as a gene) is significantly associated with the phenotype, while non-genetic information (such as demographic data) is taken into account; (2) how to deal with missing values in genotype data via the relatedness among individuals as well as the similarity among genetic variants; (3) if the phenotype is measured over time for every individual, how to take advantage of the abundant information to discover genes with time-related effects on the phenotype. To address question (1), a hypothesis test is proposed, which is proved being able to successfully detect genes already discovered being associated with a specific trait in previous studies. To address question (2), an imputation method is developed and it is shown that this method can improve the power of association tests. For the third challenge, a second hypothesis test is proposed and it is verified to be able to identify genes contributing to the pattern of a longitudinal trait.

Titre de l'écran-titre (visionné le 6 juin 2023) / Le sujet de ce mémoire concerne l'utilisation de données incomplètes qui sont utilisées à des fins d'apprentissage statistique, dans un contexte où une méthode de traitement des données manquantes a été appliquée aux données. La problématique motivant ce travail est la prédiction de l'abandon scolaire chez les étudiants collégiaux. La caractéristique principale de la non-réponse au sein de ces données est que les étudiants ayant le statut d'immigrant ont une non-réponse quasi complète pour certaines variables. À partir d'une étude de simulation répliquant le comportement des données collégiales, différentes méthodes d'imputation sont utilisées sur des jeux de données ayant différentes configurations de non-réponse. Ces données imputées sont ensuite utilisées pour entraîner des modèles d'apprentissage statistique afin d'en évaluer les performances. À partir des résultats de cette étude de simulation, les combinaisons de méthodes d'imputation et de modèles d'apprentissage statistique ayant le mieux performé au niveau des prédictions sont appliquées aux données collégiales afin de déterminer quelles méthodes d'imputation permettent d'obtenir les meilleures performances prédictives. / This thesis deals with the use of incomplete data, to which a missing data treatment has been applied, in a statistical learning problem. The issue motivating this project is the prediction of school dropout among college students. The main characteristic of non-response in these data is that students with immigrant status have non-response for almost all the variables. Based on a simulation study replicating the behavior of college data, different imputation methods are applied on datasets with different nonresponse patterns. The imputed data are then used to train statistical learning models and to evaluate their performance. Based on the results of the simulation study, the best-performing combinations of imputation methods and statistical learning models are applied to college data.

L'objectif de ce mémoire est de comparer la méthode d'échantillonnage de Gibbs avec augmentation de données, telle que présentée par Paquet (2002) et Bernier-Martel (2005), avec celle de l'imputation multiple telle que présentée par Grégoire (2004). Le critère de comparaison sera le signe des coefficients estimés. Nous travaillerons dans le contexte de bases de données indépendantes et d'un modèle linéaire à choix discret. Le modèle sera exprimé en tenant compte du choix des modes de transport des ménages de la communauté urbaine de Toronto. Pour réaliser ce projet, nous utiliserons la base de données du TTS (Transportation Tomorrow Survey) de 1986 et de 1996. Les résultats n'ont pas tous été estimés par un signe cohérent à nos attentes. Toutefois, nous pouvons conclure que l'échantillonnage de Gibbs avec augmentation de données est une approche plus intéressante que l'imputation multiple, puisqu'elle a estimé un nombre plus élevé de bons signes.

本研究主要在探討近年來兩大租稅改革：兩稅合一制及最低稅負制之施行對上市公司外資持股比率之影響，是否降低外資持有我國上市公司股票之意願，影響我國企業的股權結構。 兩稅合一制施行後，消除股利所得之重複課稅，國內法人股東與自然人實質稅負減輕；但根據所得稅法第七十三條之二規定，非中華民國境內居住之個人、在中華民國境內無固定營業場所及營業代理人之營利事業，其獲配股利所含之稅額，不得扣抵其應納稅額，外資無法享受股東可扣抵稅額消除重複課稅的好處，外資在台投資實質總稅負仍維持為40%，較境內自然人股東之實質稅率6%~40%為高，股東可扣抵稅額比例愈高，外資損失越大。再者，最低稅負制施行主要影響對象為外資持股最多的電子業，且最低稅負制之施行對於公司補繳之稅負，外資股東也無法享受股利抵稅權。故可推論兩稅合一制、最低稅負制之施行及股東可扣抵稅額與上市公司外資持股比率應呈負向關係。 本研究之樣本為民國85年到民國95年及最低稅負制前、後之台灣上市公司。主要實證結果彙整如下： 1、兩稅合一制之施行及股東可扣抵稅額與上市公司外資持股比率呈負向關係，最低稅負制之施行與上市公司外資持股比率呈正向關係，但其效果應為股價指數之影響所致而非最低稅負制之施行；又最低稅負制施行後，外資仍偏好持有公司有效稅率低於10%之股票。 2、另外以民國93年到民國95年資料額外檢測，最低稅負制下限制投資抵減金額之效用，推論投資抵減金額在最低稅負制施行後與上市公司外資持股比率應呈負向關係，實證結果為負向關係但不顯著。 / The objective of this study is to examine whether the implementation of the Imputation Tax System and Alternative Minimum Tax System will affect the willing of foreign investors to invest in listed companies. The sample of this study consists of firms listed in the Taiwan Stock Exchange from 1996 through 2006. The empirical results in this research are summarized as follows: 1、 The implementation of the Imputation Tax System and a firms' imputation tax credit ratios have a negative impact on the percentage of foreign ownership in listed companies, but the implementation of Alternative Minimum Tax System has a positive impact on that. We conduct further examination and find the increasing effect of the percentage of foreign ownership is more rely on the effect of stock index than that of the implementation of Alternative Minimum Tax System. Further, foreign investors still prefer to invest in listed companies whose effect rates are lower than 10%. 2、With the implementation of Alternative Minimum Tax System, the use of investment tax credit was restricted. The empirical result of this study shows that after the implementation of Alternative Minimum Tax System, the investment tax credit has a negative but not significant impact on the percentage of foreign ownership in listed companies.

Les données manquantes sont fréquentes dans les enquêtes et peuvent entraîner d’importantes erreurs d’estimation de paramètres. Ce mémoire méthodologique en sociologie porte sur l’influence des données manquantes sur l’estimation de l’effet d’un programme de prévention. Les deux premières sections exposent les possibilités de biais engendrées par les données manquantes et présentent les approches théoriques permettant de les décrire. La troisième section porte sur les méthodes de traitement des données manquantes. Les méthodes classiques sont décrites ainsi que trois méthodes récentes. La quatrième section contient une présentation de l’Enquête longitudinale et expérimentale de Montréal (ELEM) et une description des données utilisées. La cinquième expose les analyses effectuées, elle contient : la méthode d’analyse de l’effet d’une intervention à partir de données longitudinales, une description approfondie des données manquantes de l’ELEM ainsi qu’un diagnostic des schémas et du mécanisme. La sixième section contient les résultats de l’estimation de l’effet du programme selon différents postulats concernant le mécanisme des données manquantes et selon quatre méthodes : l’analyse des cas complets, le maximum de vraisemblance, la pondération et l’imputation multiple. Ils indiquent (I) que le postulat sur le type de mécanisme MAR des données manquantes semble influencer l’estimation de l’effet du programme et que (II) les estimations obtenues par différentes méthodes d’estimation mènent à des conclusions similaires sur l’effet de l’intervention. / Missing data are common in empirical research and can lead to significant errors in parameters’ estimation. This dissertation in the field of methodological sociology addresses the influence of missing data on the estimation of the impact of a prevention program. The first two sections outline the potential bias caused by missing data and present the theoretical background to describe them. The third section focuses on methods for handling missing data, conventional methods are exposed as well as three recent ones. The fourth section contains a description of the Montreal Longitudinal Experimental Study (MLES) and of the data used. The fifth section presents the analysis performed, it contains: the method for analysing the effect of an intervention from longitudinal data, a detailed description of the missing data of MLES and a diagnosis of patterns and mechanisms. The sixth section contains the results of estimating the effect of the program under different assumptions about the mechanism of missing data and by four methods: complete case analysis, maximum likelihood, weighting and multiple imputation. They indicate (I) that the assumption on the type of MAR mechanism seems to affect the estimate of the program’s impact and, (II) that the estimates obtained using different estimation methods leads to similar conclusions about the intervention’s effect.

A l’heure de la réforme du droit des obligations, il n’était pas inutile de revenir sur un phénomène remarqué du droit privé : l’émergence de la faute de fonction. Celle-ci interroge le privatiste quant à la possibilité de transposer dans sa matière une institution de droit administratif : la faute de service. Deux conditions doivent impérativement être remplies pour que la faute de fonction devienne une notion juridique opératoire.La première condition a pour objet de garantir que l’introduction de cette notion ne sera pas source d’insécurité juridique. Or, seule une conceptualisation de la faute de fonction pourrait permettre d’atteindre cet objectif. Celle-ci explique pourquoi la faute de fonction concerne les préposés et les dirigeants de personne morale : ces deux agents exercent communément une fonction pour le compte d’une entreprise. Ce point commun explique que leurs fautes de fonction correspondent aux mêmes critères de définition.La seconde condition a pour objet de vérifier que la faute de fonction peut être opérationnelle en droit de la responsabilité. Fondé sur la théorie du risque-profit et la théorie du risque anormal de l'entreprise, ce régime, articulé autour de la notion d’imputation, est particulièrement efficient en droit de la responsabilité civile où les fonctions de réparation et de sanction doivent être conciliées. En droit de la responsabilité pénale, droit sanctionnateur, la faute de fonction ne semble devoir s’exprimer que de façon très résiduelle. / At the time of contract law reform, it’s not unnecessary to go back to a noticed phenomenon of private law: the rise of the “misconduct within their function”. This concept raises questions for private lawyers with regards to the transferability of an administrative law concept into their own domain: the administrative fault. Two conditions must be satisfied in order for the misconduct within the function to become an operative legal concept.The first condition is about guaranteeing that introducing this concept will not be the source of legal uncertainty. However, only a conceptualisation of the “misconduct within the function” could achieve this goal. It explains why (its) liability applies to employees and leaders of a legal person: these two agents commonly undertake a task on behalf of the company. This common denominator explains that their liability equate to the same definition criteria. The second condition checks that the misconduct within their function can be operational in tort law. Based on the benefit-risk theory and the abnormal risk theory of the company, this system structured around the idea of imputation is particularly efficient in the law of civil liability where repair functions and sanctions must be reconciled. In criminal law liability, sanctioning law, the “misconduct within their function” appears to only be expressed in a residual way.