In this thesis, we explore two different approaches to causal inferences. The traditional approach models the theoretical relationship between the outcome variables and their explanatory variables, i.e., the science, at the same time as the systematic differences between treated and control subjects are modeled, i.e., the assignment mechanism. The alternative approach, based on Rubin's Causal Model (RCM), makes it possible to model the science and the assignment mechanism separately in a two-step procedure. In the first step, no outcome variables are used when the assignment mechanism is modeled, the treated students are matched with similar control students using this mechanism, and the models for the science are determined. Outcome variables are only used in the second step when these pre-specified models for the science are fitted. In the first paper, we use the traditional approach to evaluate whether a husband is more prone to quit smoking when his wife quits smoking than he would have been had his wife not quit. We find evidence that this is the case, but that our analysis must rely on restrictive assumptions. In the subsequent two papers, we use the alternative RCM approach to evaluate if a Harvard freshman who does not smoke (observed potential outcome) is more prone to start smoking when he shares a suite with at least one smoker, than he would have been had he shared a suite with only smokers (missing potential outcomes). We do not find evidence that this is the case, and the small and insignificant treatment effect is robust against various assumptions that we make regarding covariate adjustments and missing potential outcomes. In contrast, we do find such evidence when we use the traditional approach previously used in the literature to evaluate peer effects relating to smoking, but the treatment effect is not robust against the assumptions that we make regarding covariate adjustments. These contrasting results in the two latter papers allow us to conclude that there are a number of advantages with the alternative RCM approach over the traditional approaches previously used to evaluate peer effects relating to smoking. Because the RCM does not use the outcome variables when the assignment mechanism is modeled, it can be re-fit repeatedly without biasing the models for the science. The assignment mechanism can then often be modeled to fit the data better and, because the models for the science can consequently better control for the assignment mechanism, they can be fit with less restrictive assumptions. Moreover, because the RCM models two distinct processes separately, the implications of the assumptions that are made on these processes become more transparent. Finally, the RCM can derive the two potential outcomes needed for drawing causal inferences explicitly, which enhances the transparency of the assumptions made with regard to the missing potential outcomes. / Diss. Stockholm : Handelshögskolan, 2006 S. 1-13: sammanfattning, s. [15]-161: 4 uppsatser

Made available in DSpace on 2010-04-20T21:00:04Z (GMT). No. of bitstreams: 4 Angela Sayuri Cristofoli Ueno.pdf.jpg: 3172 bytes, checksum: cee83d936530b3fbd5bbd2e1e33630f0 (MD5) Angela Sayuri Cristofoli Ueno.pdf.txt: 101059 bytes, checksum: 8aed03cbd2b118345bf95e9de842bd8b (MD5) license.txt: 4712 bytes, checksum: 4dea6f7333914d9740702a2deb2db217 (MD5) Angela Sayuri Cristofoli Ueno.pdf: 1282852 bytes, checksum: f569242d1dee8eb86948ddc7774f30c9 (MD5) Previous issue date: 2010-02-08T00:00:00Z / The operational risk management and measurement is an increasing concern throughout the community of financial institutions. The adequate choice of the operational risk capital calculation model can become a competitive differential. This study presents the advantages of adopting causal models for operational risk management and measuring. The investigation of the Artificial Neural Networks application for this purpose shows that the causal model results in capital amounts more aligned to the financial institution’s risk exposure. Furthermore, there is the advantage that, as more risk sensible the capital calculation methodology is, higher will be the incentive for an appropriate risk management in the day-today institution’s business. This not only reduces the needs for capital allocation, but also decreases the expected losses. Therefore, the results are positive and encourage future researches about this subject. / A gestão e a mensuração do risco operacional é uma preocupação crescente da comunidade bancária, de modo que a escolha adequada do modelo de alocação de capital para risco operacional pode tornar-se um diferencial competitivo. Este trabalho apresenta as vantagens da adoção de modelos causais para a gestão e mensuração do risco operacional e, ao investigar a aplicação de Redes Neurais Artificiais para tal propósito, comprova que o modelo causal chega a valores de capital mais alinhados à exposição ao risco da instituição financeira. Além disso, há a vantagem de que, quanto mais sensível a risco a metodologia de cálculo de capital for, maior será o incentivo para uma gestão apropriada dos riscos no dia-a-dia da instituição financeira, o que não apenas reduz sua necessidade de alocação de capital, quanto diminui suas perdas esperadas. Os resultados, portanto, são positivos e motivam estudos futuros sobre o tema.

In order for a medication treatment to be considered successful, several roles and functions along the pharmacological treatment chain must function and cooperate effectively. The chain can most easily be described as five transitions; diagnosis, prescription treatment, dispensing, drug use and finally results and follow-up. Unfortunately, there are many problems and inefficiencies in the pharmacological drug chain. Unfortunately, those who study medication errors and their solutions have focused on individual parts of the pharmacological treatment system. However, for this reason, this study aims to develop a general framework for system analysis of inefficiencies along the pharmacological treatment chain. Due to the size of the problem, this project focused on medication adherence. Adherence can be defined as to what extent the patient follows the medication treatment plan. Adherence has many known problems and difficulties, among other things, it has major financial consequences. It can also be difficult to measure compliance, and there is no recognized perfect method. A system dynamic model is a theoretical image of a real system or object, which is a model used to understand the nonlinear behavior of complex systems. These models are useful when considering interventions and their effects when there are complex relationships. The project started with a literature study, and then went into data collection. Here, a search design and refinements were designed to find relevant articles. Once the articles were selected, the data was compiled from the articles and the analysis began. Here, factors and effects on adherence were identified as well as other interesting information from the articles. When the information was compiled and analyzed, the system dynamic model was created. The model was then sent via email to experts in the field to validation and revise the model. During the data collection, 23 relevant articles were found, compiled into 38 factors associated with compliance. In addition to these factors, 8 were excluded because they were too disease-specific or too ambiguous in their effect of adherence. The various articles studied many different chronic diseases, but hypertension was the most common. How adherence was measured in the articles also varied greatly, however, some form of self-report or questionnaire was most common method used. Three out of seven experts responded to the sent-out model and provided valuable comments. Although these are not sufficient to validate the model, their views showed that a validation can be designed in this way. The model would have to be sent to a larger set of experts and stakeholders, but because these experts are recognized in their fields, it gave weight to the results even though they were few reviewers. With the support of the literature and the experts’ statement, it was concluded that this model provides a good foundation and structure to continue to build upon. In addition, the model has proven to have many key relationships and cornerstones with important and relevant factors. It is also concluded that it is possible to translate the model into quantitative patterns, which is based on the fact that the factor itself can be translated quantitatively. Overall, it is also finally concluded that the model created in this project could be of great use in future projects when working towards a framework for system analysis of inefficiencies along the pharmacological treatment chain.

This research aims to improve the design and placement of runway exits at airports through analysis and modeling of the effects that exogenous causal factors have on pilots' landing behavior and exit selections. Incorporating these factors into modeling software will strengthen the software's utility by providing project teams the ability to specify which pilot motivational causal factors apply to a new or existing runway. The main findings suggest pilots' exit selections are deterministic but dependent on the presence (or absence) of six (6) causal factors. A model and two (2) case studies are presented and compared against predictions generated by existing modeling software. The results support a finding that the causal factor model improves motivation-based predictions over current modeling techniques, which are drawn from stochastic distributions. The accuracy of this model enables designers to optimize runway exit placement and geometry to maximize runway capacity. / Master of Science / Airport design engineers currently plan the locations and geometric characteristics of runway exits by balancing the expected fleet mix of aircraft on that runway with the capacity and delay effects that the number and placement of these exits might cause. This technique originated from research beginning in the early 1970s, which found that pilots' exit motivations primarily resulted from the capabilities and limitations of their aircraft. Since pilots tend to "fly by the numbers" (i.e., exhibit predictable approach airspeeds, power levels, wing flaps, touchdown locations, landing speeds, and braking efforts), engineers thus employed design principles in which the numbers, locations and geometries of exits were primarily functions of the physical and performance-based characteristics of the specific types of aircraft expected to utilize the runway. However, in studying more than 4 million landings by a single aircraft type (the Boeing 737-800) at 42 U.S. airports, the evidence in this thesis shows that pilots' exit selections are behaviorally motivated by more than the physics of motion. This thesis aims to refine previous research and engineering methods by showing evidence that pilots' exit selections have as much to do with the presence (or absence) of certain environmental factors within the landing system. These factors (described in detailed within) are unique to each airport's overall physical network of interconnected runways, exits, taxiways, terminals and other features. Within this network, a pilot's landing behavior and exit selection depends on the locational and relational interactions that each exit choice will have on the time and distance to their apron (gate) assignment. These "interactions" are referred to as causal factors – defined as physical features within a landing environment that pilots have little-to-no control over – but which nevertheless influence their specific exit selections. Two (2) runway case studies provided in this thesis evidence a finding that a causal factor model reliably predicts pilots' exit selections better than current modeling techniques, which are drawn from probability-based statistical distributions. The stability and accuracy of the new model enables engineering design and project teams to optimize runway exit placement and geometry to maximize runway capacity, and can be adopted for use in both existing and future runways.

This doctoral thesis is devoted to the study of the effects of late fertility on health of the newborn and to the implications of late fertility in the fertility decision process. Late fertility is defined as the reproduction process after 30 years old. The interest lies as well from the maternal age point of view as from the paternal age point of view. The first part is devoted to the study of the determinants of infant morbidity and mortality with a particular attention to the parental age, without neglecting the other determinants. The originality of this part is located from the methodological point of view. We construct a structural model of infant morbidity/mortality in order to isolate the causal effect of late fertility. By a structural model we mean a model which represents a set of causal relationships represented mathematically by a multi-equation model and graphically by directed acyclic graphs. As a complementary approach, a chapter of the thesis is devoted to an exploratory model highlighting the role of the extreme values rather than average values traditionally of interest in most statistical analyses. Both methods are tested with Hungarian data: individual registration forms of livebirths and infant deaths (1984-1984 and 1994-1998), and the Hungarian case-control surveillance of congenital abnormalities (1997-2002). The second part analyses the effect of parental ageing in the fertility decision process. We try to determine whether the detrimental effect of late fertility on health of the child and on fecundity of the couples intervene in the preferences for a child. We assume that parental age influences the preferences for a child through effects on the desire for a child and on the beliefs in the capacity of reproduction of a healthy child. This hypothesis is tested using the data from the National Survey of Family Growth (United States, 2002) and from the Fertility and Family Survey (Hungary, Czech Republic and Belgium).

Dans ce mémoire par articles, nous nous intéressons à l’apprentissage de modèles causaux à partir de données. L’intérêt de cette entreprise est d’obtenir une meilleure compréhension des données et de pouvoir prédire l’effet qu’aura un changement sur certaines variables d’un système étudié. Comme la découverte de liens causaux est fondamentale en sciences, les méthodes permettant l’apprentissage de modèles causaux peuvent avoir des applications dans une pléthore de domaines scientifiques, dont la génomique, la biologie et l’économie. Nous présentons deux nouvelles méthodes qui ont la particularité d’être des méthodes non-linéaires d’apprentissage de modèles causaux qui sont posées sous forme d’un problème d’optimisation continue sous contrainte. Auparavant, les méthodes d’apprentissage de mo- dèles causaux abordaient le problème de recherche de graphes en utilisant des stratégies de recherche voraces. Récemment, l’introduction d’une contrainte d’acyclicité a permis d’abor- der le problème différemment. Dans un premier article, nous présentons une de ces méthodes: GraN-DAG. Sous cer- taines hypothèses, GraN-DAG permet d’apprendre des graphes causaux à partir de données observationnelles. Depuis la publication du premier article, plusieurs méthodes alternatives ont été proposées par la communauté pour apprendre des graphes causaux en posant aussi le problème sous forme d’optimisation continue avec contrainte. Cependant, aucune de ces méthodes ne supportent les données interventionnelles. Pourtant, les interventions réduisent le problème d’identifiabilité et permettent donc l’utilisation d’architectures neuronales plus expressives. Dans le second article, nous présentons une autre méthode, DCDI, qui a la particularité de pouvoir utiliser des données avec différents types d’interventions. Comme le problème d’identifiabilité est moins important, une des deux instanciations de DCDI est un approximateur de densité universel. Pour les deux méthodes proposées, nous montrons que ces méthodes ont de très bonnes performances sur des données synthétiques et réelles comparativement aux méthodes traditionelles. / In this thesis by articles, we study the learning of causal models from data. The goal of this entreprise is to gain a better understanding of data and to be able to predict the effect of a change on some variables of a given system. Since discovering causal relationships is fundamental in science, causal structure learning methods have applications in many fields that range from genomics, biology, and economy. We present two new methods that have the particularity of being non-linear methods learning causal models casted as a continuous optimization problem subject to a constraint. Previously, causal strutural methods addressed this search problem by using greedy search heuristics. Recently, a new continuous acyclity constraint has allowed to address the problem differently. In the first article, we present one of these non-linear method: GraN-DAG. Under some assumptions, GraN-DAG can learn a causal graph from observational data. Since the publi- cation of this first article, several alternatives methods have been proposed by the community by using the same continuous-constrained optimization formulation. However, none of these methods support interventional data. Nevertheless, interventions reduce the identifiability problem and allow the use of more expressive neural architectures. In the second article, we present another method, DCDI, that has the particularity to leverage data with several kinds of interventions. Since the identifiabiliy issue is less severe, one of the two instantia- tions of DCDI is a universal density approximator. For both methods, we show that these methods have really good performances on synthetic and real-world tasks comparatively to other classical methods.

L'intérêt en apprentissage machine pour étudier la causalité s'est considérablement accru ces dernières années. Cette approche est cependant encore peu répandue dans le domaine de l’intelligence artificielle (IA) et du droit. Elle devrait l'être. L'approche associative actuelle d’apprentissage machine révèle certaines limites que l'analyse causale peut surmonter. Cette thèse vise à découvrir si les modèles causaux peuvent être utilisés en IA et droit. Nous procédons à une brève revue sur le raisonnement et la causalité en science et en droit. Traditionnellement, les cadres normatifs du raisonnement étaient la logique et la rationalité, mais la théorie duale démontre que la prise de décision humaine dépend de nombreux facteurs qui défient la rationalité. À ce titre, des statistiques et des probabilités étaient nécessaires pour améliorer la prédiction des résultats décisionnels. En droit, les cadres de causalité ont été définis par des décisions historiques, mais la plupart des modèles d’aujourd’hui de l'IA et droit n'impliquent pas d'analyse causale. Nous fournissons un bref résumé de ces modèles, puis appliquons le langage structurel de Judea Pearl et les définitions Halpern-Pearl de la causalité pour modéliser quelques décisions juridiques canadiennes qui impliquent la causalité. Les résultats suggèrent qu'il est non seulement possible d'utiliser des modèles de causalité formels pour décrire les décisions juridiques, mais également utile car un schéma uniforme élimine l'ambiguïté. De plus, les cadres de causalité sont utiles pour promouvoir la responsabilisation et minimiser les biais. / The machine learning community’s interest in causality has significantly increased in recent years. This trend has not yet been made popular in AI & Law. It should be because the current associative ML approach reveals certain limitations that causal analysis may overcome. This research paper aims to discover whether formal causal frameworks can be used in AI & Law. We proceed with a brief account of scholarship on reasoning and causality in science and in law. Traditionally, normative frameworks for reasoning have been logic and rationality, but the dual theory has shown that human decision-making depends on many factors that defy rationality. As such, statistics and probability were called for to improve the prediction of decisional outcomes. In law, causal frameworks have been defined by landmark decisions but most of the AI & Law models today do not involve causal analysis. We provide a brief summary of these models and then attempt to apply Judea Pearl’s structural language and the Halpern-Pearl definitions of actual causality to model a few Canadian legal decisions that involve causality. Results suggest that it is not only possible to use formal causal models to describe legal decisions, but also useful because a uniform schema eliminates ambiguity. Also, causal frameworks are helpful in promoting accountability and minimizing biases.

Les analyses effectuées dans le cadre de ce mémoire ont été réalisées à l'aide du module MatchIt disponible sous l’environnent d'analyse statistique R. / Statistical analyzes of this thesis were performed using the MatchIt package available in the statistical analysis environment R. / L’estimation sans biais de l’effet causal d’une intervention nécessite la comparaison de deux groupes homogènes. Il est rare qu’une étude observationnelle dispose de groupes comparables et même une étude expérimentale peut se retrouver avec des groupes non comparables. Les chercheurs ont alors recours à des techniques de correction afin de rendre les deux groupes aussi semblables que possible. Le problème consiste alors à choisir la méthode de correction appropriée. En ce qui nous concerne, nous limiterons nos recherches à une famille de méthodes dites d’appariement. Il est reconnu que ce qui importe lors d’un appariement est l’équilibre des deux groupes sur les caractéristiques retenues. Autrement dit, il faut que les variables soient distribuées de façon similaire dans les deux groupes. Avant même de considérer la distribution des variables entre les deux groupes, il est nécessaire de savoir si les données en question permettent une inférence causale. Afin de présenter le problème de façon rigoureuse, le modèle causal contrefactuel sera exposé. Par la suite, les propriétés formelles de trois méthodes d’appariement seront présentées. Ces méthodes sont l’appariement par la distance de Mahalanobis, de l’appariement par le score de propension et de l’appariement génétique. Le choix de la technique d’appariement appropriée reposera sur quatre critères empiriques dont le plus important est la différence des moyennes standardisées. Les résultats obtenus à l’aide des données de l’Enquête longitudinale et expérimentale de Montréal (ÉLEM) indiquent que des trois techniques d’appariement, l’appariement génétique est celui qui équilibre mieux les variables entre les groupes sur tous les critères retenus. L’estimation de l’effet de l’intervention varie sensiblement d’une technique à l’autre, bien que dans tous les cas cet effet est non significatif. Ainsi, le choix d’une technique d’appariement influence l’estimation de l’effet d’une intervention. Il est donc impérieux de choisir la technique qui permet d’obtenir un équilibre optimal des variables selon les données à la disposition du chercheur. / The unbiased estimate of the causal effect of an intervention requires the comparison of two homogeneous groups. It is rare that an observational study has comparable groups and even an experiment may end up with non-comparable groups. The researchers then used correction techniques to make the two groups as similar as possible. The problem then is to choose the appropriate correction method. In our case, we will restrict our research to a family of so-called matching methods. It is recognized that what matters in a match is the balance between the two groups on selected characteristics. In other words, it is necessary that the variables are distributed similarly in both groups. Even before considering the distribution of variables between the two groups, it is necessary to know whether the data in question allow for causal inference. To present the problem rigorously, the counterfactual causal model will be exposed. Thereafter, the formal properties of three matching methods will be presented. Those methods are the Mahalanobis matching, the propensity score matching and genetic matching. The choice of the appropriate matching technique is based on four empirical criteria which the most important is the standardized mean difference. Results obtained using data from the Montréal Longitudinal and Experimental Study indicate that of the three matching techniques, genetic matching is the one that better balance the variables between groups on all criteria. The estimate of the effect of intervention varies substantially from one technique to another, although in all cases this effect is non significant. Thus, the selection of a matching technique influences the estimation of the effect of an intervention. Therefore, it is imperative to choose the technique that provides an optimal balance of the variables based on data available to the researcher.

本研究以「人口遷移學推拉理論」為基礎，探索我國與留學地主國間留學互動關係的推拉因果模式，及其間的一致性和關聯性，並對我國未來留學生人數進行預測。研究的主要目的為：(1)瞭解我國與留學地主國留學教育概況，並探討海外留學對留學地主國與送出留學生國家國家的正反面效果；(2)說明人口遷移學中的推拉理論及評閱有關留學生流動的研究文獻與報告；(3)探索我國與留學地主國間留學互動的推拉因果模式；(4)探索我國與留學地主國間在留學互動的推拉因果模式間之一致性與關聯性 ；(5)對我國未來出國的留學生人數進行預測；最後(6)根據研究結果提出建議，作為制訂留學教育政策及未來研究的參考。 　　在探索留學互動的推拉因果模式與其模式間的一致性和關聯性時，是以我國與美國、德國、日本及法國等四個留學地主國為對象，採1954年到1988年共35年縱貫動態分析為主。在我國未來留學生人數預測上，則以1950年到1988年的動態資料為主。研究資料來源是「中華民國教育部統計」、「中華民國台灣地區統計提要」、「中華民國統計年鑑」、UNESCO統計、國際貨幣基金統計年報、美國國際教育組織的Open doors統計，做為分析的根據。 　　本研究之資料處理係利用國立政治大學PRIME 6150大電腦的SPSSX、SAS/ETS及PC版的LISREL 7統計套裝軟體，另外引用余民寧(民81)所設計「二次式分配準則SAS/IML之程式」，作為統計分析的工具。本研究共提出十個虛無假設，並擬以下列方法檢定研究假設。 　　一、以共變結構分析(LISREL)檢定我國與留學地主國間留學互動的推拉因果模式，即假設一～四。 　　二、以二次式分配準則 (QAP)檢定我國與留學地主國在留學互動的推拉因果模式之一致性與關聯性，即假設五～十。 　　三、以單變量時間數列ARIMA方法與迴歸分析方法，進行我國未來留學生人數之預測。 　　本研究之主要結果為： 　　一、我國與美國間留學互動的推拉因果關係證實存在。 　　二、我國與德國間留學互動的推拉因果關係證實存在。 　　三、我國與日本間留學互動的推拉因果關係證實存在。 　　四、我國與法國間留學互動的推拉因果關係在修正模式後證實存在。 　　五、我國與美國、我國與德國在留學互動的推拉因果模式之適配共變數矩陣具有.429的顯著相關性與一致性。 　　六、我國與美國、我國與法國在留學互動的推拉因果模式之適配共變數矩陣具有.469的顯著相關性與一致性。 　　七、我國與美國、我國與日本在留學互動的推拉因果模式之適配共變數矩陣的相關性與一致性僅-.098而已。 　　八、我國與德國、我國與法國在留學互動的推拉因果模式之適配共變數矩陣具有.763的顯著相關性與一致性。 　　九、我國與德國、我國與日本在留學互動的推拉因果模式之適配共變數矩陣具有.510的顯著相關性與一致性。 　　十、我國與法國、我國與日本在留學互動的推拉因果模式之適配共變數矩陣具有.377的顯著相關性與一致性。 　　另外，在我國未來出國留學人數預測上，民國87年以前預期每年將至少有6600名以上的留學生出國，並且當國民所得達12000美元時，出國留學的人數預期將可能突破10000人以上。 　　本研究根據研究結果提出建議，作為政府制訂留學教育政策及未來研究的參考。 / This research is based on "the push-pull theory of population mobility. It explores between Taiwan and the leading host coun-tries the causal model, consistency and correlation of the push-pull interaction in studying abroad. It also forecasts the number of studying-abroad students of Taiwan in the future. Therefore, the purposes of this research are: (1) to understand the foreign education of both Taiwan and the leading host coun-tries and further to probe the pros and cons of foreign educa-tion; (2) to explain the push-pull theory of population mobility and to comment the literatures of studying abroad; (3) to explore between Taiwan and the leading host countries the causal rela-tionship of push-pull interaction in studying abroad; (4) to explore between Taiwan and the leading host countries the consistency and correlation of push-pull causal model in studying abroad; (5) to forecast the number of studying abroad students of Taiwan; and (6) to propose suggestions for the policy-making of studying abroad and future studies according to the results of this research. 　　In exploring the causal relationship model, consistency, and correlation of the push-pull interaction in studying abroad, the subjects will be Taiwan, U.S.A., Germany, Japan, and France. The data are collected from The R.O.C. St-atis-bics of the Educa-tion Ministry, The R.O.C. Statistics Summary of Taiwan Areas, The R.O.C. Statistics Yearbook, UNESCO Statistical Yearbook, Interna-tional Financial Statistics Yearbook, and Open Doors (1991-1993) of the Institute of International Education. While in forecast-ing the number of studying-abroad students of Taiwan the data will be ranged from 1950 to 1988. All data of this research are dynamic. 　　The handling of data will adopt SPSSX, SAS/ETS, and LISREL7 packages program and will cite Yu Min-ning"s SAS/IML program of QAP (1992). All packages program are in the Computer Center (PRIME 6150) of National Cheng-chi University, exclusive of LIS-REL7 which is set in personal computer. This research will propose ten null hypotheses, and the statistical methods used to confirm the null hypotheses are as follows: 　　(1) Use Linear Struc-tural Equation (LISREL) to test the causal relationship of the push-pull interaction in studying abroad between Taiwan and the leading host countries. (Hypotheses 1-4) 　　(2) Use Quadratic Assignment Paradigm (QAP) to test the con-sistency, correlation of the push-pull interaction in studying abroad between Taiwan and leading host countries. (Hypotheses 5-10) 　　(3) Use both Autoregerssion Integrated Moving Average (ARIMA) of univarate time series and regression analysis to forecast the number of the studying-abroad students of Taiwan in the future. 　　The main results of this research are as follows: 　　(1) There exists a push-pull causal' relationship in studying abroad between Taiwan and U. S. A. . 　　(2) There exists a push-pull causal relationship in studying abroad between Taiwan and Germany. 　　(3) There exists a push-pull causal relationship in studying abroad between Taiwan and Japan. 　　(4) There exists a push-pull causal relationship in studying abroad between Taiwan and France after modifying the model. 　　(5) Taiwan-U.S.A. and Taiwan-Germany best-fitted covariance matrices are significantly similar. The correlation coefficient is .429. 　　(6) Taiwan-U.S.A. and Taiwan-France best-fitted covariance matrices are significantly similar. The correlation coefficient 　　(7) Taiwan-U.S.A and Taiwan-Japan best-fitted covariance matrices are not significantly similar. The correlation coeffi-cient is only -.098. 　　(8) Taiwan-Germany and Taiwan-France best-fitted covariance matrices are significantly similar. The correlation coefficient is .763. 　　(9) Taiwan-Germany and Taiwan-Japan best-fitted covariance ma-trices are significantly similar. The correlation coefficient is .510. 　　(10) Taiwan-France and Taiwan-Japan best-fitted covariance ma-trices are significantly similar. The correlation coefficient is .3768. 　　Therefore, nine null hypotheses are rejected and only one null hypothesis is accepted. 　　Besides, in forecasting the number of the studying-abroad students of Taiwan, it will be expected to send out over 6600 students to study abroad every year before 1998. Furthermore, when the per capita income of Taiwan reaches US$12000, the number of studying-abroad students will be over 10000 per year. 　　Finally, according to conclusions and results of this re-search, some suggestions for the policy-making of studying abroad and future studies in this field are proposed.