Causal inference on networks often encounters interference problems. The potentialoutcomes of a unit depend not only on its treatment but also on the treatments of its neighbors in the network. The classic causal inference assumption of no interference among units is untenable in networks, and many fundamental results in causal inference may no longer hold in the presence of interference. To address interference problems in networks, this thesis proposes a novel Network Embedding Matching (NEM) framework for estimating causal effects under network interference. We recover causal effects based on network structure in an observed network. Furthermore, we extend the network interference from direct neighbors to k-hop neighbors. Unlike most previous studies, which had strong assumptions on interference among units in the network and did not consider network structure, our framework incorporates network structure into the estimation of causal effects. In addition, our NEM framework can be implemented in networks for randomized experiments and observational studies. Our approach is interpretable and can be easily applied to networks. We compare our approach with other existing methods in simulations and real networks, and we show that our approach outperforms other methods under linear and nonlinear network interference. / Statistics

Autistic traits are a composition of behavioral constructs that encompasses social functioning, communication, and rigid and repetitive behaviors that might impact an individual’s quality of life. The specificity of these traits is not yet fully understood, nor which traits that might be most debilitating for autistic people. We recruited 366 participants, out of which 78 were diagnosed as autistic, and measured levels of different character traits as well as their quality of life. We ran a Bayesian regression model and found extreme evidence that the behavioral constructs of prosopagnosia, social anhedonia, alexithymia and cognitive empathy contribute to autistic social functioning, while affective empathy did not seem to contribute to the same extent. To estimate the effect of each construct on quality of life we employed Causal Inference methodology and found likely effects of social anhedonia (-0.131 [-0.248, 0.00]) and alexithymia (-0.255 [-0.37, -0.154]). Therefore, both social anhedonia and alexithymia might be effective targets for intervention for autistic people struggling with social functioning.

<p>Mediation analyses play a critical role in social and personality psychology research. However, current approaches for assessing power and sample size in mediation models have limitations, particularly when dealing with complex mediation models and multiple mediator sequential models. These limitations stem from limited software options and the substantial computational time required. In this part, we address these challenges by extending the joint significance test and product of coefficients test to incorporate the fourth-pathed mediated effect and generalized kth-pathed mediated effect. Additionally, we propose a model-based bootstrap method and provide convenient R tools for estimating power in complex mediation models. Through our research, we demonstrate that power decreases as the number of mediators increases and as the influence of coefficients varies. We summarize our results and discuss the implications of power analysis in relation to mediator complexity and coefficient influence. We provide insights for researchers seeking to optimize study designs and enhance the reliability of their findings in complex mediation models. </p> <p>Matching is a crucial step in causal inference, as it allows for more robust and reasonable analyses by creating better-matched pairs. However, in real-world scenarios, data are often collected and stored by different local institutions or separate departments, posing challenges for effective matching due to data fragmentation. Additionally, the harmonization of such data needs to prioritize privacy preservation. In this part, we propose a new hierarchical framework that addresses these challenges by implementing differential privacy on raw data to protect sensitive information while maintaining data utility. We also design a data access control system with three different access levels for designers based on their roles, ensuring secure and controlled access to the matched datasets. Simulation studies and analyses of datasets from the 2017 Atlantic Causal Inference Conference Data Challenge are conducted to showcase the flexibility and utility of our framework. Through this research, we contribute to the advancement of statistical methodologies in matching and privacy-preserving data analysis, offering a practical solution for data integration and privacy protection in causal inference studies. </p> <p>Biomarker discovery is a complex and resource-intensive process, encompassing discovery, qualification, verification, and validation stages prior to clinical evaluation. Streamlining this process by efficiently identifying relevant biomarkers in the discovery phase holds immense value. In this part, we present a likelihood ratio-based approach to accurately identify truly relevant protein markers in discovery studies. Leveraging the observation of unimodal underlying distributions of expression profiles for irrelevant markers, our method demonstrates promising performance when evaluated on real experimental data. Additionally, to address non-normal scenarios, we introduce a kernel ratio-based approach, which we evaluate using non-normal simulation settings. Through extensive simulations, we observe the high effectiveness of the kernel method in discovering the set of truly relevant markers, resulting in precise biomarker identifications with elevated sensitivity and a low empirical false discovery rate.  </p>

京都大学 / 新制・課程博士 / 博士(情報学) / 甲第23542号 / 情博第772号 / 新制||情||132(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 山本 章博, 教授 下平 英寿 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

Depression increases the risk of suicide death and non-fatal suicide attempt. Between 2 - 6% of persons with depression will die by suicide1 and 25 - 31% of persons with depression will make a non-fatal suicide attempt during their lifetime.2,3 Despite the strong association between depression and suicidal behavior, the vast majority of persons with depression will not engage in suicidal behavior, making it difficult to accurately predict who is at risk for suicide and non-fatal suicide attempt. Identifying high risk persons who should be connected to suicide prevention interventions is an important public health goal. Furthermore, depression often co-occurs with other mental disorders, which may exert an interactive influence on the risk of suicide and suicide attempt. Understanding the joint influence of depression and other mental disorders on suicide outcomes may inform prevention strategies. The goals of this dissertation were to predict suicide and non-fatal suicide attempt among persons with depression and to quantify the causal joint effect of depression and comorbid psychiatric disorders on suicide and suicide attempt. For all three studies, we used data from Danish registries, which routinely collect high-quality data in a setting of universal health care with long-term follow-up and registration of most health and life events.4 In Study 1, we predicted suicide deaths among men and women diagnosed with depression using a case-cohort design (n = 14,737). Approximately 800 predictors were included in the machine learning models (classification trees and random forests), spanning demographic characteristics, income, employment, immigrant status, citizenship, family suicidal history (parent or spouse), previous suicide attempts, mental disorders, physical health disorders, surgeries, prescription drugs, and psychotherapy. In depressed men, we found interactions between hypnotics and sedatives, analgesics and antipyretics, and previous poisonings that were associated with a high risk of suicide. In depressed women, there were interactions between poisoning and anxiolytics and between anxiolytics and hypnotics and sedatives that were associated with suicide risk. The variables in the random forests that contributed the most to prediction accuracy in depressed men were previous poisoning diagnoses and prescriptions of hypnotics and sedatives and anxiolytics. In depressed women, the most important predictors of suicide were receipt of state pension, prescriptions for psychiatric medications (anxiolytics and antipsychotics) and diagnoses of poisoning, alcohol related disorders, and reaction to severe stress and adjustment disorders. Prescriptions of analgesics and antipyretics (e.g., acetaminophen) and antithrombotic agents (e.g., aspirin) emerged as important predictors for both depressed men and women. Study 2 predicted non-fatal suicide attempts among men and women diagnosed with depression using a case-cohort design (n = 17,995). Among depressed men, there was a high risk of suicide attempt among those who received a state pension and were diagnosed with toxic effects of substances. There was also an interaction between reaction to severe stress and adjustment disorder and not receiving psychological help that was associated with suicide attempt risk among depressed men. In depressed women, suicide attempt risk was high in those who were prescribed antipsychotics, diagnosed with specific personality disorders, did not have a poisoning diagnosis, and were not receiving a state pension. For both men and women, the random forest results showed that the strongest contributors to prediction accuracy of suicide attempts were poisonings, alcohol related disorders, reaction to severe stress and adjustment disorders, drugs used to treat psychiatric disorders (e.g., drugs used in addictive disorders, anxiolytics, hypnotics and sedatives), anti-inflammatory medications, receipt of state pension, and remaining single. Study 3 examined the joint effect of depression and other mental disorders on suicide and non-fatal suicide attempts using a case-cohort design (suicide death analysis n = 279,286; suicide attempt analysis n = 288,157). We examined pairwise combinations of depression with: 1) organic disorders, 2) substance use disorders, 3) schizophrenia, 4) bipolar disorder, 5) neurotic disorders, 6) eating disorders, 7) personality disorders, 8) intellectual disabilities, 9) developmental disorders, and 10) behavioral disorders. We fit sex-stratified joint marginal structural Cox models to account for time-varying confounding. We observed large hazard ratios for the joint effect of depression and comorbid mental disorders on suicide and suicide attempts, the effect of depression in the absence of comorbid mental disorders, and for the effect of comorbid mental disorders in the absence of depression. We observed positive and negative interdependence between different combinations of depression and comorbid mental disorders on the rate of suicide and suicide attempt, with variation by sex. Overall, depression and comorbid mental disorders are harmful exposures, both independently and jointly. All of the studies in this dissertation highlight the important role of interactions between risk factors in suicidal behavior among persons with depression. Depression is one of the most commonly assessed risk factors for suicide,5,6 and our findings underscore the value of considering additional risk factors such as other psychiatric disorders, psychiatric medications, and social factors in combination with depression. The results of this dissertation may help inform potential risk identification strategies which may facilitate the targeting of suicide prevention interventions to those most vulnerable.

Thesis advisor: Laura O'Dwyer / Black males are disproportionately placed in special education throughout the United States. Yet, the degree to which such disparities are warranted has been subject to debate. Prior research suggests that special education is used too often in high-poverty schools partly due to limited resources available to support struggling students (Skiba et al., 2006). More recent studies, however, suggest that, when considering student background characteristics and peer racial and socioeconomic composition, Black students are underrepresented in special education, specifically in high-minority schools (Elder et al., 2021). Given these varying findings and interpretations, in this dissertation I sought to bring further clarity to the issue of disproportionality as it relates to Black males. First, I replicated previous research using student-level data from two high-poverty school districts based in a Northeastern state to examine variation in special education placement by race and gender, before and after adjusting for background characteristics. To then understand whether special education placement was effective, I used student fixed effect models to estimate how academic achievement trajectories changed for students after placement and whether these findings differed by race and gender. I found that Black males in the sample were placed in special education at higher rates than students of other race-by-gender groups, even after adjusting for background characteristics. Prior to placement, Black males experienced large declines in academic achievement, and this trend continued after receiving special education. Together, these findings support the notion that Black males are likely overrepresented in special education. Provided these findings, in the second part of this dissertation, I tested the effectiveness of a potential policy mechanism in reducing disproportionality. Specifically, I asked whether providing teachers with additional resources to direct struggling students through a comprehensive student support program reduced the probability of special education placement for Black males. Using two distinct identification strategies, I found that this form of support reduced special education placement rates for Black students, nearly eliminating their disproportionate representation in the districts. I conclude with policy implications for both measuring and addressing disproportionality. / Thesis (PhD) — Boston College, 2022. / Submitted to: Boston College. Lynch School of Education. / Discipline: Educational Research, Measurement and Evaluation.

Cette mémoire vise à développer une méthode de pondération par l’inverse de le probabilité de traitement (Augmented Inverse Probability of Treatment Weighting; AIPTW) et estimation par maximum de vraisemblance ciblée (Targeted Maximum Likelihood Estimation; TMLE) dans le contexte d'une méta-analyse en réseau avec données individuelles (Individual Patient Data Network Meta-Analysis; IPD-NMA) avec données observationnelles. Nous proposons également des méthodes pour estimer le score de propension généralisé (Generalized Propensity Score; GPS) pour finalement estimer l'effet causal d'une combinaison donnée de traitements (un régime) interprété à partir de d’une population globale. Cette recherche a été motivée par une mise à jour récente des données de patients atteints de la tuberculose multirésistante (Multidrug-Resistant Tuberculosis ; MDR-TB), une maladie infectieuse respiratoire causée par le bacillus mycobactérie avec un taux de mortalité élevé. Une compléxité notable de notre scénario est que toutes les régimes de traitements n'ont pas été observés dans toutes les études. L’inférence causale est définie comme l'étude de l'effet des traitements sur un résultat. Bien que les études cliniques randomisées sont l'étalon-or pour l'investigation des causes et effets, en raison de certaines limitations, leur utilisation n'est pas toujours faisable. Ainsi, l’analyse de données observationnelles est proposée. Donc, il est important de développer des méthodes qui nous permettent d'utiliser les informations provenant des données observationnelles. L'utilisation des informations provenant de plusieurs études individuelles nous permet d'évaluer les associations entre les traitements et les résultats qui sont spécifiques aux sous-populations. Aussi, une méta-analyse en réseau nous permet comparer plusieurs régimes au lieu de seulement deux. Nous estimons le taux de succès d’un régime donné à partir d'un ensemble d'études dans lesquelles le régime était disponible, puis le généralisons à l'ensemble de la population source. La théorie et les résultats d’une étude de simulation démontre que les méthodes développées sont doublement robustes. Cependant, TMLE démontre plus de robustesse, en particulier lorsqu’une méthode nouvellement proposée pour estimer le GPS est utilisée. Le résultat de l'application donne des estimations d’un taux de succès de traitement généralisé entre 50 à 61 % pour le régime {Pyrazinamide,Kanamycin,Ofloxacin,Ethionamide,Cyloserine} tandis que le taux observé de l’ensemble des données était de 59 %. / This thesis aims for developing Augmented Inverse Probability of Treatment Weighting (AIPTW) and Targeted Maximum Likelihood Estimation (TMLE) in the setting of Individual Patient Data Network Meta-Analysis (IPD-NMA) of observational data and propose a method to estimate the Generalized Propensity Score (GPS) to eventually estimate the causal effect of a given combination of treatments (a regimen) and generalize it to a global population. This research was motivated by a recent update on IPD_NMA of Multidrug-Resistant Tuberculosis (MDR-TB) - a respiratory infectious disease caused by bacillus mycobacterium with a high rate of mortality - where not all the regimens observed in all the studies. Although Randomized Controlled Trials (RCTs) are known to be the gold standard in investigating cause-and-effect including in causal inference (defined as the study of the effect of treatments on an outcome), but because of some known limitations using them is not always feasible. Thus, observational data are being proposed. Therefore, developing methods that enable us to use the information from observational data is important. In addition, using the information coming from individual studies allows us to evaluate associations between treatments and outcome which are specific to subpopulations. Also, a network meta-analysis allows us to study the effect of multiple treatments instead of two. We estimate the rate of treatment success for a given regimen from a set of studies where the regimen was available, and then generalize it to the whole network. The simulation result shows that the developed methods are doubly robust, however TMLE shows more robustness specially when the new proposed approach to estimate the GPS is being used. The application result shows a range of 50-61% for the generalized success rate of regimen {Pyrazinamide,Kanamycin,Ofloxacin,Ethionamide,Cyloserine} while the observed rate was 59% from multiple regimens.

The present study provides an example of implementing the difference in differences (DD) estimator for a two-group, pretest-posttest design with K-12 educational intervention data. The goal is to explore the basis for causal inference via Rubin's potential outcomes framework. The DD method is introduced to educational researchers, as it is seldom implemented in educational research. DD analytic methods' mathematical formulae and assumptions are explored to understand the opportunity and the challenges of using the DD estimator for causal inference in educational research. For this example, the teacher intervention effect is estimated with multi-cohort student outcome data. First, the DD method is used to detect the average treatment effect (ATE) with linear regression as a baseline model. Second, the analysis is repeated using linear regression with cluster robust standard errors. Finally, a linear mixed effects analysis is provided with a random intercept model. Resulting standard errors, parameter estimates, and inferential statistics are compared among these three analyses to explore the best holistic analytic method for this context.

Bipartite experiments have recently emerged as a focal point in causal inference. In these experiments, treatment is administered to one set of units, while outcomes of interest are gauged on a distinct set of units. Such experiments are especially valuable in scenarios where pronounced interference effects transpire between units on a bipartite network. For instance, in market experiments, designating treatment at the seller level and assessing outcomes at the buyer level (or vice-versa) can lead to causal models that more accurately reflect the inherent interference between buyers and sellers. Although bipartite experiments can enhance the precision of causal effect estimations in specific contexts, it's imperative to conduct the analysis judiciously to avoid introducing undue bias through the network. Drawing from the generalized propensity score literature, we demonstrate that it's feasible to achieve unbiased estimates of causal effects for bipartite experiments, given a conventional set of assumptions. Furthermore, we delve into the formulation of confidence sets with accurate coverage probabilities. By employing a bipartite graph from a publicly accessible dataset previously explored in bipartite experiment studies, we illustrate, via simulations, a notable reduction in bias and augmented coverage. / Statistics

No description available.