Univ., Diss.--St. Gallen, 2001. / Includes bibliographical references.

Indiana University-Purdue University Indianapolis (IUPUI) / Observational data are frequently used for causal inference of treatment effects on prespecified outcomes. Several widely used causal inference methods have adopted the method of inverse propensity score weighting (IPW) to alleviate the in uence of confounding. However, the IPW-type methods, including the doubly robust methods, are prone to large variation in the estimation of causal e ects due to possible extreme weights. In this research, we developed an ordinary least-squares (OLS)-based causal inference method, which does not involve the inverse weighting of the individual propensity scores. We first considered the scenario of homogeneous treatment effect. We proposed a two-stage estimation procedure, which leads to a model-free estimator of average treatment effect (ATE). At the first stage, two summary scores, the propensity and mean scores, are estimated nonparametrically using regression splines. The targeted ATE is obtained as a plug-in estimator that has a closed form expression. Our simulation studies showed that this model-free estimator of ATE is consistent, asymptotically normal and has superior operational characteristics in comparison to the widely used IPW-type methods. We then extended our method to the scenario of heterogeneous treatment effects, by adding in an additional stage of modeling the covariate-specific treatment effect function nonparametrically while maintaining the model-free feature, and the simplicity of OLS-based estimation. The estimated covariate-specific function serves as an intermediate step in the estimation of ATE and thus can be utilized to study the treatment effect heterogeneity. We discussed ways of using advanced machine learning techniques in the proposed method to accommodate high dimensional covariates. We applied the proposed method to a case study evaluating the effect of early combination of biologic & non-biologic disease-modifying antirheumatic drugs (DMARDs) compared to step-up treatment plan in children with newly onset of juvenile idiopathic arthritis disease (JIA). The proposed method gives strong evidence of significant effect of early combination at 0:05 level. On average early aggressive use of biologic DMARDs leads to around 1:2 to 1:7 more reduction in clinical juvenile disease activity score at 6-month than the step-up plan for treating JIA.

京都大学 / 新制・課程博士 / 博士(情報学) / 甲第24727号 / 情博第815号 / 新制||情||137(附属図書館) / 京都大学大学院情報学研究科知能情報学専攻 / (主査)教授 鹿島 久嗣, 教授 阿久津 達也, 教授 下平 英寿 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

In clinical trials, placebo response is considered a beneficial effect arising from multiple factors, including the patient’s expectations for the treatment. Due to the presence of placebo response, the classical parallel design often fails to declare an efficacious treatment. The Sequential Parallel Comparison Design (SPCD), a two-stage design where the first stage is a classical parallel trial, followed by another parallel trial among placebo patients from the first stage, was proposed to mitigate the placebo response. In SPCD, in lieu of treatment effect, a weighted average of the mean treatment difference in Stage I among all randomized patients and the mean treatment difference in Stage II among placebo non-responders was proposed as the efficacy measure. However, by mixing two possibly different populations, this weighted average lacks interpretability, the choice of weight remains controversial, and the classification of patients into placebo responders and non-responders via hard criterion-based rule warrants careful consideration. In this work, we first elaborate and study the shortcomings surrounding this efficacy measure, which motivates us to propose causal estimands for clinically meaningful principal strata under the principal stratification framework. To make the estimands identifiable, we invoke a set of assumptions and introduce two sensitivity parameters. Meanwhile, in the absence of a clinically proven criterion for classifying responders and non-responders, we additionally suggest estimating the response status and sensitivity parameters via the Expectation-Maximization (EM) algorithm by treating the principal strata as full latent characteristics. Next, we further refine and alternatively propose a more consistent and sophisticated EM procedure for classification, point estimation, and hypothesis testing. Finally, we evaluate our methods with extensive simulation studies and apply them to an actual SPCD study of antidepressant therapy to assess the benefit of low-dose aripiprazole adjunctive to antidepressant therapy treatment, the ADAPT-A trial. In conclusion, we believe this is an important step toward a more rigorous and transparent approach to evaluating the treatment benefit in the presence of placebo response. / 2025-09-18T00:00:00Z

The focus of this research is on hypotheses testing involving inequality constraints. In the first chapter of this dissertation, we propose Kolmogorov-Smirnov type tests for stochastic dominance relations between the potential outcomes of a binary treatment under the unconfoundedness assumption. Our stochastic dominance tests compare every point of the cumulative distribution functions (CDF), so they can fully utilize all information in the distributions. For first order stochastic dominance, the test statistic is defined as the supremum of the difference of two inverse-probability-weighting estimators for the CDFs of the potential outcomes. The critical values are approximated based on a simulation method. We show that our test has good size properties and is consistent in the sense that it can detect any violation of the null hypothesis asymptotically. First order stochastic dominance tests in the treated subpopulation, and higher order stochastic dominance tests in the whole population and among the treated are shown to share the same properties. The tests are applied to evaluate the effect of a job training program on incomes, and we find that job training has a positive effect on real earnings. Finally, we extend our tests to cases in which the unconfoundedness assumption does not hold. On the other hand, there has been a considerable amount of attention paid to testing inequality restrictions using Wald type tests. As noted by Wolak (1991), there are certain situations where it is difficult to obtain tests with correct size even asymptotically. These situations occur when the variance-covariance matrix of the functions in the constraints depends on the unknown parameters as would be the case in nonlinear models. This dependence on the unknown parameters makes it computationally difficult to find the least favorable configuration (LFC) which can be used to bound the size of the test. In the second chapter of this dissertation, we extend Hansen's (2005) superior predictive ability (SPA) test to testing hypotheses involving general inequality constraints in which the variance-covariance matrix can be dependent on the unknown parameters. For our test we are able to obtain correct size asymptotically plus test consistency without requiring knowledge of the LFC. Also the test can be applied to a wider class of problems than considered in Wolak (1991). In the last chapter, we construct new Kolmogorov-Smirnov tests for stochastic dominance of any pre-specified order without resorting to the LFC to improve the power of Barrett and Donald's (2003) tests. To do this, we first show that under the null hypothesis if the objects being compared at a given income level are not equal, then the objects at this given income level will have no effect on the null distribution. Second, we extend Hansen's (2005) recentering method to a continuum of inequality constraints and construct a recentering function that will converge to the underlying parameter function uniformly asymptotically under the null hypothesis. We treat the recentering function as a true underlying parameter function and add it to the simulated Brownian bridge processes to simulate the critical values. We show that our tests can control the size asymptotically and are consistent. We also show that by avoiding the LFC, our tests are less conservative and more powerful than Barrett and Donald's (2003). Monte Carlo simulations support our results. We also examine the performances of our tests in an empirical example. / text

Antonín Hanuš 5. prosince 2014 This work deals with randomized clinical trials of medicaments. It examines three models of dependece of final values on initial values in case, that all variables are measured with some measurement error. For each model is derived effect of treatment estimate and its asymptotical properties, specifically consistency and asymptotical variance. The work mostly deals with linear model of analysis of covariance ANCOVA. The work fruther contents comparison of properties of estimates from all three models in case, that examined data come from a linear model. There is a comparison of asymptotical variances of estimates from all three models and for each of them there are examined conditions, when this model gives the best results. In the end there is a simulation study, which verifies all previous results. 1

Thesis advisor: Henry I. Braun / City Connects is a school-based model that identifies the strengths and needs of every student and links each child to a tailored set of intervention, prevention, and enrichment services in the school or community. The purpose of this study was to conduct a comprehensive evaluation of the City Connects treatment effects on academic performance (both MCAS scores and grade point average (GPA) grades) in middle school using student longitudinal records. Parallel analyses were conducted: one evaluated the City Connects elementary intervention (serving kindergarten to fifth grades) and the other one evaluated the City Connects middle school intervention (serving sixth to eighth grades). A series of two-level hierarchical linear models with middle school achievement scores adjusted and/or propensity score weights applied were used to answer the research questions of interest. In addition, to make a causal inference, a sensitivity analysis was conducted to examine whether or not the estimated treatment effects resulted from the first two analyses were robust to the presence of unobserved selection bias. The results showed that students who were exposed to the City Connects elementary intervention significantly outperformed their counterparts, who graduated from the comparison elementary schools, on academic achievement in all middle school grades. However, in the case of the City Connects intervention schools that served middle school grades, since all students only received a maximum of one year of City Connects middle school intervention, it was still too soon to expect any significant changes. Moreover, the estimated treatment effects of the City Connects elementary intervention were only mildly sensitive to the presence of some forms of hidden bias, which made the causal inference of City Connects on middle school academic achievement quite plausible. / Thesis (PhD) — Boston College, 2015. / Submitted to: Boston College. Lynch School of Education. / Discipline: Educational Research, Measurement and Evaluation.

Os modelos de transição de Markov constituem uma ferramenta de grande importância para diversas áreas do conhecimento quando são desenvolvidos estudos com medidas repetidas. Eles caracterizam-se por modelar a variável resposta ao longo do tempo condicionada a uma ou mais respostas anteriores, conhecidas como a história do processo. Além disso, é possível a inclusão de outras covariáveis. No caso das respostas binárias, pode-se construir uma matriz com as probabilidades de transição de um estado para outro. Neste trabalho, quatro abordagens diferentes de modelos de transição foram comparadas para avaliar qual estima melhor o efeito causal de tratamentos em um estudo experimental em que a variável resposta é um vetor binário medido ao longo do tempo. Estudos de simulação foram realizados levando em consideração experimentos balanceados com três tratamentos de natureza categórica. Para avaliar as estimativas foram utilizados o erro padrão, viés e percentual de cobertura dos intervalos de confiança. Os resultados mostraram que os modelos de transição marginalizados são mais indicados na situação em que um experimento é desenvolvido com um reduzido número de medidas repetidas. Como complementação, apresenta-se uma forma alternativa de realizar comparações múltiplas, uma vez que os pressupostos como normalidade, independência e homocedasticidade são violados impossibilitando o uso dos métodos tradicionais. Um experimento com dados reais no qual se registrou a presença de fungos (considerada como sucesso) em cultivos de citros e morango foi analisado por meio do modelo de transição apropriado. Para as comparações múltiplas, intervalos de confiança simultâneos foram construídos para o preditor linear e os resultados foram estendidos para a resposta média que neste caso são as probabilidades de sucesso. / The transition Markov models are a very important tool for several areas of knowledge when studies are developed with repeated measures. They are characterized by modeling the response variable over time conditional to the previous response which is known as the history. In addtion it is possible to include other covariates. In the case of binary responses, can be constructed a matrix of transition probabilities from one state to another. In this work, four different approaches to transition models were compared in order to assess which best estimates of the causal effect of treatments in an experimental studies where the outcome is a vector of binary response measured over time. Simulation study was held taking into account a balanced experiments with three treatments of categorical nature. To assess the best estimates standard error and bias, beyond the percentage of coverage were used. The results showed that the marginalized transition models are more appropriate in situation where an experiment is developed with a reduced number of repeated measurements. As complementation is presented an alternative way to perform multiple comparisons, since the assumptions as normality, independence and homoscedasticity are violated precluding the use of traditional methods. An experiment with real data where we recorded the presence of fungi (deemed successful) in citrus and strawberry crops was analyzed through the appropriate transition model. For multiple comparisons, simultaneous confidence intervals were constructed for the linear predictor and the results have been extended to the mean response in this case are the probabilities of success.

Comparative experiments, in which subjects are randomized to one of two treatments, are performed often. There is no shortage of papers testing whether a treatment effect exists and providing confidence intervals for the magnitude of this effect. While it is well understood that the object and scope of inference for an experiment will depend on what assumptions are made, these entities are not always clearly presented. We have proposed one possible method, which is based on the ideas of Jerzy Neyman, that can be used for constructing confidence intervals in a comparative experiment. The resulting intervals, referred to as Neyman-type confidence intervals, can be applied in a wide range of cases. Special care is taken to note which assumptions are made and what object and scope of inference are being investigated. We have presented a notation that highlights which parts of a problem are being treated as random. This helps ensure the focus on the appropriate scope of inference. The Neyman-type confidence intervals are compared to possible alternatives in two different inference settings: one in which inference is made about the units in the sample and one in which inference is made about units in a fixed population. A third inference setting, one in which inference is made about a process distribution, is also discussed. It is stressed that certain assumptions underlying this third type of inference are unverifiable. When these assumptions are not met, the resulting confidence intervals may cover their intended target well below the desired rate. Through simulation, we demonstrate that the Neyman-type intervals have good coverage properties when inference is being made about a sample or a population. In some cases the alternative intervals are much wider than necessary on average. Therefore, we recommend that researchers consider using our Neyman-type confidence intervals when carrying out inference about a sample or a population as it may provide them with more precise intervals that still cover at the desired rate.

Chronic diseases place a substantial economic burden on the health care system. Physical inactivity, poor diet and smoking are considered to be the main causes of high rates of chronic disease. Evidence clearly supports the positive influence of physical activity on health determinants, other health outcomes and quality of life. This implies that an increase in physical activity improves general health status and has the potential to reduce utilization of expensive healthcare services and disability days. Earlier studies show that physical activity programs would be an effective way of providing preventive care for people with chronic conditions. However studies that relate physical activity programs to health care utilization are limited in economics literature.<p>The aim of this paper is to examine the impact of physical activity programs on healthcare utilization. From 2002 to 2003, adults over the age of 50 years, in a mid-size Canadian city, presenting with excess weight, type 2 diabetes, hypertension, hyperlipidemia or osteoarthritis were recruited. Following a screening process, eligible participants were randomly assigned to one of two programs: a class-based structured program or a home-based unstructured program. Validated questionnaires related to health status and quality of life were completed and physical tests were carried out at baseline, 3, 6, 12 months and 24 months after the program initiation. In addition participants use of physician and hospital services and pharmaceutical expenditures were accessed through their administrative data files for three years, one year before and two years after the intervention. Using administrative data from Sask Health and individual level survey data the effects of physical activity programs on health care utilization were estimated. The results showed that structured physical activity program can reduce annual physician costs significantly. The exponential effect of aging was found to be significant on hospital utilization, and the number of comorbidities was found to be significant on prescription drug utilization.