arXiv subjects

Daniel Nevo

Publications and source records attributed to Daniel Nevo.

At least 19 recordsLinked to original sources

arXiv2026

Sensitivity analysis for contamination in egocentric-network randomized trials with interference

Egocentric-Network Randomized Trials (ENRTs) are increasingly used to estimate causal effects under interference when measuring complete sociocentric network data is infeasible. ENRTs rely on egocentric network sampling, where a set of egos is first sampled, and each ego recruits a subset of its neighbors as alters. Treatments are then randomized across egos. While the observed ego-networks are disjoint by design, the underlying population network may contain edges connecting them, leading to contamination. Under a design-based framework, we show that the Horvitz-Thompson estimators of direct and indirect effects are biased whenever contamination is present. To address this, we derive bias-corrected estimators and propose a novel sensitivity analysis framework based on sensitivity parameters representing the probability or expected number of missing edges. This framework is implemented via both grid sensitivity analysis and probabilistic bias analysis, providing researchers with a flexible tool to assess the robustness of the causal estimators to contamination. We apply our methodology to the HIV Prevention Trials Network 037 study, finding that ignoring contamination may lead to underestimation of indirect effects and overestimation of direct effects.

stat.ME↗

arXiv2025

Nonparametric bounds for vaccine effects in randomized trials

Vaccine randomized trials are typically designed to be blinded, ensuring that the estimated vaccine efficacy (VE) reflects the immunological effect of the vaccine. When blinding is broken, however, the estimated VE reflects not only the immunological effect but also behavioral effects stemming from participants' awareness of their treatment status. Recent work has proposed alternative causal estimands to the standard VE to address this issue, but their point identification results require a strong assumption: the absence of unmeasured common causes of infection risk and participants' belief about whether they received the vaccine. Personality traits, for example, may plausibly violate this assumption. We relax this assumption and derive nonparametric causal bounds for different types of VE. We construct these bounds using two approaches: linear programming-based and monotonicity-based methods. We further consider several possible causal structures for vaccine trials and show how the nonparametric bounds differ across these scenarios. Finally, we illustrate the performance of the proposed bounds using fully synthetic data and a semi-synthetic data example based on a COVID-19 vaccine trial.

stat.ME↗

arXiv2025

Set Valued Predictions For Robust Domain Generalization

Despite the impressive advancements in modern machine learning, achieving robustness in Domain Generalization (DG) tasks remains a significant challenge. In DG, models are expected to perform well on samples from unseen test distributions (also called domains), by learning from multiple related training distributions. Most existing approaches to this problem rely on single-valued predictions, which inherently limit their robustness. We argue that set-valued predictors could be leveraged to enhance robustness across unseen domains, while also taking into account that these sets should be as small as possible. We introduce a theoretical framework defining successful set prediction in the DG setting, focusing on meeting a predefined performance criterion across as many domains as possible, and provide theoretical insights into the conditions under which such domain generalization is achievable. We further propose a practical optimization method compatible with modern learning architectures, that balances robust performance on unseen domains with small prediction set sizes. We evaluate our approach on several real-world datasets from the WILDS benchmark, demonstrating its potential as a promising direction for robust domain generalization.

cs.LG↗

arXiv2025

Causal effects on non-terminal event time with application to antibiotic usage and future resistance

Comparing future antibiotic resistance levels resulting from different antibiotic treatments is challenging because some patients may survive only under one of the antibiotic treatments. We embed this problem within a semi-competing risks approach to study the causal effect on resistant infection, treated as a non-terminal event time. We argue that existing principal stratification estimands for such problems exclude patients for whom a causal effect is well-defined and is of clinical interest. Therefore, we present a new principal stratum, the infected-or-survivors (ios). The ios is the subpopulation of patients who would have survived or been infected under both antibiotic treatments. This subpopulation is more inclusive than previously defined subpopulations. We target the causal effect among these patients, which we term the feasible-infection causal effect (FICE). We develop large-sample bounds under novel assumptions, and discuss the plausibility of these assumptions in our application. As an alternative, we derive FICE identification using two illness-death models with a bivariate frailty random variable. These two models are connected by a cross-world correlation parameter. Estimation is performed by an expectation-maximization algorithm followed by a Monte Carlo procedure. We apply our methods to detailed clinical data obtained from a hospital setting.

stat.ME↗

arXiv2025

Bayesian Estimation of Causal Effects Using Proxies of a Latent Interference Network

Network interference occurs when treatments assigned to some units affect the outcomes of others. Traditional approaches often assume that the observed network correctly specifies the interference structure. However, in practice, researchers frequently only have access to proxy measurements of the interference network due to limitations in data collection or potential mismatches between measured networks and actual interference pathways. In this paper, we introduce a framework for estimating causal effects when only proxy networks are available. Our approach leverages a structural causal model that accommodates diverse proxy types, including noisy measurements, multiple data sources, and multilayer networks, and defines causal effects as interventions on population-level treatments. The latent nature of the true interference network poses significant challenges. To overcome them, we develop a Bayesian inference framework. We propose a Block Gibbs sampler with Locally Informed Proposals to update the latent network, thereby efficiently exploring the high-dimensional posterior space composed of both discrete and continuous parameters. The latent network updates are driven by information from the proxy networks, treatments, and outcomes. We illustrate the performance of our method through numerical experiments, demonstrating its accuracy in recovering causal effects even when only proxies of the interference network are available.

stat.ME↗

arXiv2024

Correcting invalid regression discontinuity designs with multiple time period data

Regression Discontinuity (RD) designs rely on the continuity of potential outcome means at the cutoff, but this assumption often fails when other treatments or policies are implemented at this cutoff. We characterize the bias in sharp and fuzzy RD designs due to violations of continuity, and develop a general identification framework that leverages multiple time periods to estimate local effects on the (un)treated. We extend the framework to settings with carry-over effects and time-varying running variables, highlighting additional assumptions needed for valid causal inference. We propose an estimation framework that extends the conventional and bias-corrected single-period local linear regression framework to multiple periods and different sampling schemes, and study its finite-sample performance in simulations. Finally, we revisit a prior study on fiscal rules in Italy to illustrate the practical utility of our approach.

econ.EM↗

arXiv2023

Negative Control Falsification Tests for Instrumental Variable Designs

The validity of instrumental variable (IV) designs is typically tested using two types of falsification tests. We characterize these tests as conditional independence tests between negative control variables -- proxies for unobserved variables posing a threat to the identification -- and the IV or the outcome. We describe the conditions that variables must satisfy in order to serve as negative controls. We show that these falsification tests examine not only independence and the exclusion restriction, but also functional form assumptions. Our analysis reveals that conventional applications of these tests may flag problems even in valid IV designs. We offer implementation guidance to address these issues.

econ.EM↗

arXiv2023

Distinguishing immunological and behavioral effects of vaccination

The interpretation of vaccine efficacy estimands is subtle, even in randomized trials designed to quantify immunological effects of vaccination. In this article, we introduce terminology to distinguish between different vaccine efficacy estimands and clarify their interpretations. This allows us to explicitly consider immunological and behavioural effects of vaccination, and establish that policy-relevant estimands can differ substantially from those commonly reported in vaccine trials. We further show that a conventional vaccine trial allows identification and estimation of different vaccine estimands under plausible conditions, if one additional post-treatment variable is measured. Specifically, we utilize a ``belief variable'' that indicates the treatment an individual believed they had received. The belief variable is similar to ``blinding assessment'' variables that are occasionally collected in placebo-controlled trials in other fields. We illustrate the relations between the different estimands, and their practical relevance, in numerical examples based on an influenza vaccine trial.

stat.AP↗

arXiv2023

Learn-As-you-GO (LAGO) Trials: Optimizing Treatments and Preventing Trial Failure Through Ongoing Learning

It is well known that changing the intervention package while a trial is ongoing does not lead to valid inference using standard statistical methods. However, it is often necessary to adapt, tailor, or tweak a complex intervention package in public health implementation trials, especially when the intervention package does not have the desired effect. This article presents conditions under which the resulting analyses remain valid even when the intervention package is adapted while a trial is ongoing. Our results on such Learn-As-you-GO (LAGO) studies extend the theory of LAGO for binary outcomes following a logistic regression model (Nevo, Lok and Spiegelman, 2021) to LAGO for continuous outcomes under flexible conditional mean model. We derive point and interval estimators of the intervention effects and ensure the validity of hypothesis tests for an overall intervention effect. We develop a confidence set for the optimal intervention package, which achieves a pre-specified mean outcome while minimizing cost, and confidence bands for the mean outcome under all intervention package compositions. This work will be useful for the design and analysis of large-scale intervention trials where the intervention package is adapted, tailored, or tweaked while the trial is ongoing.

stat.ME↗

arXiv2023

Causal inference with misspecified network interference structure

Under interference, the treatment of one unit may affect the outcomes of other units. Such interference patterns between units are typically represented by a network. Correctly specifying this network requires identifying which units can affect others -- an inherently challenging task. Nevertheless, most existing approaches assume that a known and accurate network specification is given. In this paper, we study the consequences of such misspecification. We derive bounds on the bias arising from estimating causal effects using a misspecified network, showing that the estimation bias grows with the divergence between the assumed and true networks, quantified through their induced exposure probabilities. To address this challenge, we propose a novel estimator that leverages multiple networks simultaneously and remains unbiased if at least one of the networks is correct, even when we do not know which one. Therefore, the proposed estimator provides robustness to network specification. We illustrate key properties and demonstrate the utility of our proposed estimator through simulations and analysis of a social network field experiment.

stat.ME↗

arXiv2022

The Subtype-Free Average Causal Effect for Heterogeneous Disease Etiology

Studies have shown that the effect an exposure may have on a disease can vary for different subtypes of the same disease. However, existing approaches to estimate and compare these effects largely overlook causality. In this paper, we study the effect smoking may have on having colorectal cancer subtypes defined by a trait known as microsatellite instability (MSI). We use principal stratification to propose an alternative causal estimand, the Subtype-Free Average Causal Effect (SF-ACE). The SF-ACE is the causal effect of the exposure among those who would be free from other disease subtypes under any exposure level. We study non-parametric identification of the SF-ACE, and discuss different monotonicity assumptions, which are more nuanced than in the standard setting. As is often the case with principal stratum effects, the assumptions underlying the identification of the SF-ACE from the data are untestable and can be too strong. Therefore, we also develop sensitivity analysis methods that relax these assumptions. We present three different estimators, including a doubly-robust estimator, for the SF-ACE. We implement our methodology for data from two large cohorts to study the heterogeneity in the causal effect of smoking on colorectal cancer with respect to MSI subtypes.

stat.ME↗

arXiv2022

A sensitivity analysis approach for the causal hazard ratio in randomized and observational studies

The Hazard Ratio (HR) is often reported as the main causal effect when studying survival data. Despite its popularity, the HR suffers from an unclear causal interpretation. As already pointed out in the literature, there is a built-in selection bias in the HR, because similarly to the truncation by death problem, the HR conditions on post-treatment survival. A recently proposed alternative, inspired by the Survivor Average Causal Effect (SACE), is the causal HR, defined as the ratio between hazards across treatment groups among the study participants that would have survived regardless of their treatment assignment. We discuss the challenge in identifying the causal HR and present a sensitivity analysis identification approach in randomized controlled trials utilizing a working frailty model. We further extend our framework to adjust for potential confounders using inverse probability of treatment weighting. We present a Cox-based and a flexible non-parametric kernel-based estimation under right censoring. We study the finite-sample properties of the proposed estimation method through simulations. We illustrate the utility of our framework using two real-data examples.

stat.ME↗

arXiv2021

Matching methods for truncation by death problems

Even in a carefully designed randomized trial, outcomes for some study participants can be missing, or more precisely, ill-defined, because participants had died prior to date of outcome collection. This problem, known as truncation by death, means that the treated and untreated are no longer balanced with respect to covariates determining survival. Therefore, researchers often utilize principal stratification and focus on the Survivor Average Causal Effect (SACE). The SACE is the average causal effect among the subpopulation that will survive regardless of treatment status. In this paper, we present matching-based methods for SACE identification and estimation. We provide an identification result for the SACE that motivates the use of matching to restore the balance among the survivors. We discuss various practical issues, including the choice of distance measures, possibility of matching with replacement, and post-matching crude and model-based SACE estimators. Simulation studies and data analysis demonstrate the flexibility of our approach. Because the cross-world assumptions needed for SACE identification can be too strong, we also present sensitivity analysis techniques and illustrate their use in real data analysis. Finally, we show how our approach can also be utilized to estimate conditional separable effects, a recently-proposed alternative for the SACE.

stat.ME↗

arXiv2020

Causal inference for semi-competing risks data

An emerging challenge for time-to-event data is studying semi-competing risks, namely when two event times are of interest: a non-terminal event time (e.g. age at disease diagnosis), and a terminal event time (e.g. age at death). The non-terminal event is observed only if it precedes the terminal event, which may occur before or after the non-terminal event. Studying treatment or intervention effects on the dual event times is complicated because for some units, the non-terminal event may occur under one treatment value but not under the other. Until recently, existing approaches (e.g., the survivor average causal effect) generally disregarded the time-to-event nature of both outcomes. More recent research focused on principal strata effects within time-varying populations under Bayesian approaches. In this paper, we propose alternative non time-varying estimands, based on a single stratification of the population. We present a novel assumption utilizing the time-to-event nature of the data, which is weaker than the often-invoked monotonicity assumption. We derive results on partial identifiability, suggest a sensitivity analysis approach, and give conditions under which full identification is possible. Finally, we present non-parametric and semi-parametric estimation methods for right-censored data.

stat.ME↗

arXiv2020

Modeling semi-competing risks data as a longitudinal bivariate process

The Adult Changes in Thought (ACT) study is a long-running prospective study of incident all-cause dementia and Alzheimer's disease (AD). As the cohort ages, death (a terminal event) is a prominent competing risk for AD (a non-terminal event), although the reverse is not the case. As such, analyses of data from ACT can be placed within the semi-competing risks framework. Central to semi-competing risks, and in contrast to standard competing risks, is that one can learn about the dependence structure between the two events. To-date, however, most methods for semi-competing risks treat dependence as a nuisance and not a potential source of new clinical knowledge. We propose a novel regression-based framework that views the two time-to-event outcomes through the lens of a longitudinal bivariate process on a partition of the time scale. A key innovation of the framework is that dependence is represented in two distinct forms, $\textit{local}$ and $\textit{global}$ dependence, both of which have intuitive clinical interpretations. Estimation and inference are performed via penalized maximum likelihood, and can accommodate right censoring, left truncation and time-varying covariates. The framework is used to investigate the role of gender and having $\ge$1 APOE-$\epsilon4$ allele on the joint risk of AD and death.

stat.ME↗

arXiv2020

Causal Mediation Analysis for Interpreting Neural NLP: The Case of Gender Bias

Common methods for interpreting neural models in natural language processing typically examine either their structure or their behavior, but not both. We propose a methodology grounded in the theory of causal mediation analysis for interpreting which parts of a model are causally implicated in its behavior. It enables us to analyze the mechanisms by which information flows from input to output through various model components, known as mediators. We apply this methodology to analyze gender bias in pre-trained Transformer language models. We study the role of individual neurons and attention heads in mediating gender bias across three datasets designed to gauge a model's sensitivity to gender bias. Our mediation analysis reveals that gender bias effects are (i) sparse, concentrated in a small part of the network; (ii) synergistic, amplified or repressed by different components; and (iii) decomposable into effects flowing directly from the input and indirectly through the mediators.

cs.CL↗

arXiv2018

Analysis of "Learn-As-You-Go" (LAGO) Studies

In learn-as-you-go (LAGO) adaptive studies, the intervention is a complex package consisting of multiple components, and is adapted in stages during the study based on past outcome data. This design formalizes standard practice, and desires for practice, in public health intervention studies. An effective intervention package is sought, while minimizing intervention package cost. When analyzing data from a learn-as-you-go study, the interventions in later stages depend upon the outcomes in the previous stages, violating standard statistical theory. We develop methods for estimating the intervention effects in a LAGO study. We prove consistency and asymptotic normality using a novel coupling argument, ensuring the validity of the test for the hypothesis of no overall intervention effect. We develop a confidence set for the optimal intervention package and confidence bands for the success probabilities under alternative package compositions. We illustrate our methods in the BetterBirth Study, which aimed to improve maternal and neonatal outcomes among 157,689 births in Uttar Pradesh, India through a complex, multi-component intervention package.

stat.ME↗

arXiv2018

A novel calibration framework for survival analysis when a binary covariate is measured at sparse time points

The goals in clinical and cohort studies often include evaluation of the association of a time-dependent binary treatment or exposure with a survival outcome. Recently, several impactful studies targeted the association between aspirin-taking and survival following colorectal cancer diagnosis. Due to surgery, aspirin-taking value is zero at baseline and may change its value to one at some time point. Estimating this association is complicated by having only intermittent measurements on aspirin-taking. Naive, commonly-used, methods can lead to substantial bias. We present a class of calibration models for the distribution of the time of status change of the binary covariate. Estimates obtained from these models are then incorporated into the proportional hazard partial likelihood in a natural way. We develop nonparametric, semiparametric and parametric calibration models, and derive asymptotic theory for the methods that we implement in the aspirin and colorectal cancer study. Our methodology allows to include additional baseline variables in the calibration models for the status change time of the binary covariate. We further develop a risk-set calibration approach that is more useful in settings in which the association between the binary covariate and survival is strong.

stat.AP↗