From March 15-18, 2026, several NC State Statistics faculty, students, and postdoctoral fellows presented at this year’s International Biometric Society Eastern North American Region (ENAR) 2026 Spring Meeting in Indianapolis, Indiana. Below lists the respective titles and abstracts from this year’s participants.

Three students were recognized there with the ENAR Distinguished Student Paper Award:

• Xiaodan Zhou (Drs. Brian Reich and Shu Yang, advisors) for her work, “Spatial Causal Tensor Completion for Multiple Exposures and Outcomes”;
• Si Hyung Park (Dr. Shu Yang, advisor) for his work, “Semiparametric Proximal Difference-in-Differences for Long-Term Treatment Effect Estimation Under Confounding and Missingness”; and
• Sohyeon Kim (Drs. Wenbin Lu and Luo Xiao, advisors) for her work, “Multivariate Multinomial Logit Model with ANOVA Decomposition for Correlated Categorical Outcomes”.

Congratulations to all!

Speaker: Shih-Ni Prim 

Title: A Spectral Confounder Adjustment for Spatial Regression with Multiple Exposures and Outcomes

Authors: Shih-Ni Prim, Shu Yang, Brian J Reich and others.

Abstract: Unmeasured spatial confounding complicates exposure effect estimation in environmental health studies. This problem is exacerbated in studies with multiple health outcomes and environmental exposure variables, as the source and magnitude of confounding bias may differ across exposure/outcome pairs. We propose to mitigate the effects of spatial confounding in multivariate studies by projecting to the spectral domain to separate relationships by the spatial scale and assuming that the confounding bias dissipates at more local scales. Under this assumption and some reasonable conditions, the random effect is uncorrelated with the exposures in local scales, ensuring causal interpretation of the regression coefficients. Our model for the exposure effects is a three-way tensor over exposure, outcome, and spatial scale. We use a canonical polyadic decomposition and shrinkage priors to encourage sparsity and borrow strength across the dimensions of the tensor. We demonstrate the performance of our method in an extensive simulation study and data analysis to understand the relationship between disaster resilience and the incidence of chronic diseases.

Speaker: Sihyung Park

Title: Semiparametric Proximal Difference-in-Differences for Long-Term Treatment Effect Estimation Under Confounding and Missingness

Authors: Si Hyung Park, Shu Yang and other

Abstract: Estimating the long-term treatment effect is crucial for developing a complete understanding of a treatment’s clinical profile. The limited follow-up duration of typical randomized controlled trials (RCTs), however, presents a significant challenge. A common strategy is to supplement RCT data with external real-world cohorts. Still, existing methods for this integration rely on fully observed data and strong, untestable exchangeability assumptions, which limit their validity, particularly under common data-generating processes such as linear factor models. This work develops a novel semiparametric data integration methodology inspired by proximal causal inference. Our framework successfully integrates heterogeneous data sources even in the presence of time-varying latent confounding and informative missingness. The proposed estimator is efficient and robust for the long-term average treatment effect in settings where traditional approaches are compromised by their rigid assumptions.

Speaker: Nowrin Nusrat

Title: Landmark Competing Risk Framework for Dynamic Absolute Risk Prediction of Cognitive Decline

Authors: Nowrin Nusrat, Wenbin Lu, Jung-Ying Tzeng 

Abstract: Identifying high risk individuals before the onset of cognitive decline (CD) is critical for Alzheimer’s disease prevention. Existing genetic absolute risk (AR) models use baseline covariates and relative risk estimation, limiting dynamic interpretability. We propose a landmark competing risk framework that offers methodological guidance for developing dynamic AR models integrating genetic and time-varying non-genetic factors. The landmark cause-specific hazard (L-CSH) and landmark proportional subdistribution hazard (L-PSH) models are introduced. Both models use last value carry forward, mean, and functional principal component analysis to summarize longitudinal covariates. The framework facilitates time-varying effect modification between genetic and environmental factors and recalibration of AR at multiple landmark ages. We apply both methods to dynamically predict age specific AR of CD using the Health and Retirement Study data and evaluate performance using time-dependent AUC and Brier score. The proposed framework offers a structured, interpretable guideline for genetic AR prediction with longitudinal data, bridging methodological rigor and clinical relevance.

Speaker: Matthew Singer

Title: Masked Symptom Prediction Using NLP

Authors: Matthew Singer, Srijan Sengupta, Karl Pazdernik 

Abstract: Misreporting of patient symptoms is a persistent challenge in medical research, with prior remedies relying on manual review, imputation, or deep learning methods such as autoencoders, GANs, and CNNs. We address this issue by integrating conformal prediction with adapted symptom embeddings, inspired by NLP word embeddings. Reformulating the masked language problem as the masked symptom problem, we demonstrate that unordered symptom data can be embedded into high-dimensional vectors using modified embedding methods. Compared with pretrained biomedical language models, our embeddings perform better on the masked-symptom task and, under conformal prediction, generate smaller prediction sets on average. These results highlight embedding adaptations and masked symptom modeling as effective tools for identifying unreported patient symptoms. We encourage practitioners with access to larger, more comprehensive patient datasets to further explore the masked symptom problem.

Speaker: Ke Zhu

Title: Robust Estimation and Inference in Small Hybrid Controlled Trials with Conformal Selective Borrowing

Authors: Ke Zhu, Jiajun Liu, Shu Yang and others.

Abstract: Hybrid controlled trials (HCTs) combine randomized controlled trials (RCTs) with external real-world data to enhance efficiency, but concerns arise when external controls are not fully comparable to trial participants. We present conformal selective borrowing, a framework that integrates modern machine learning techniques with classical randomization inference. This approach transparently assesses the comparability of each external control, automatically tunes the degree of borrowing, and preserves finite-sample type I error control through randomization tests, thereby addressing a key barrier to the clinical adoption of external data. We demonstrate applications to continuous, binary, and time-to-event outcomes, and show how machine learning inspired methods can provide credible and regulatory-relevant evidence even when RCTs are small or patient accrual is slow.

Speaker: Yujing Gao

Title: Designing Efficient Randomized and Single-Arm Trials: External Control Borrowing and Sample Size Calculation

Authors: Yujing Gao, Shu Yang and others.

Abstract: External controls (ECs) from real-world data are increasingly used to augment randomized controlled trials (RCTs) and single-arm trials (SATs), especially when balanced randomization is infeasible. While most existing work has focused on post-trial inference with ECs, their role in prospective trial design remains less explored. We propose a unified design framework encompassing standard RCTs and EC-integrated designs, focusing on sample size determination and power analysis. Building on semiparametrically efficient estimators, we develop hybrid and single-arm designs that leverage comparable ECs to substantially reduce required RCT sample sizes. We derive asymptotic variances of these estimators in interpretable, population-level quantities and introduce a pre-experimental variance estimation procedure to guide sample size calculation, ensuring prespecified type I error and power. Simulations demonstrate the proposed EC-integrated designs maintain valid type I error and achieve target power across diverse scenarios while requiring fewer subjects than RCT-only designs. A real data application illustrates the practical utility and advantages of the proposed EC-integrated designs.

Speaker: Xiaodan Zhou

Title: Spatial Causal Tensor Completion for Multiple Exposures and Outcomes

Authors: Xiaodan Zhou, Brian J Reich, Shu Yang 

Abstract: Estimating causal effects with multiple exposures and multiple outcomes is especially challenging when unmeasured spatial confounding is present. We propose a spatial causal tensor completion framework that jointly models multiple exposures and outcomes within a low-rank tensor structure, while adjusting for latent spatial confounders. This method combines a low-rank tensor representation to pool information across exposures and outcomes with a spectral adjustment step that incorporates graph-Laplacian eigenvectors to approximate unmeasured spatial confounders, implemented via a projected-gradient descent algorithm. This framework enables causal inference in the presence of unmeasured spatial confounding and pervasive missingness of potential outcomes. We establish theoretical guarantees for the estimator and evaluate its finite-sample performance through extensive simulations. In an application to national PFAS monitoring data, our approach yields more conservative and credible causal relationships between PFOA/PFOS exposure and 13 chronic disease outcomes compared with existing alternatives.

Speaker: Sohyeon Kim

Title: Multivariate Multinomial Logit Model with ANOVA Decomposition for Correlated Categorical Outcomes

Authors: Sohyeon Kim, Wenbin Lu, Luo Xiao and others.

Abstract: In many medical studies, it is often to collect multiple categorical outcomes that are interdependent. However, most existing research on multinomial regression fits each outcome separately, which ignores correlations between outcomes, and thus may lead to loss of information and reduced predictive accuracy. Accounting for correlations between multilevel categorical outcomes requires a high-dimensional parameter space, making model estimation challenging. In this paper, we propose a multivariate multinomial logit model that captures outcome correlations and reduces the parameter space using ANOVA decomposition. The ANOVA decomposition enables explicit conditional model formulations, which allows for a computationally much simpler composite likelihood approach. We develop an efficient Minorization-Maximization (MM) algorithm to maximize the composite likelihood, which also incorporates variable selection via a bridge penalty. Simulation studies are conducted to evaluate our method, demonstrating its effectiveness in parameter estimation and variable selection. We further illustrate our method using data from the AURORA study.

Speaker: Shubhajit Sen 

Title: Optimal Use of Survey Weights for Causal Inference under Informative Sampling

Authors: Shubhajit Sen, Shu Yang 

Abstract: The use of survey weights in causal inference is debated when treatment assignment is ignorable at the population level, yet the sampling is informative and dependent on the outcome, given treatment and covariates. The literature lacks consensus on where to incorporate survey weights: in the causal estimator, propensity score, outcome regression, or all stages, limiting guidance for researchers. We provide theoretical justification for using weights at each stage to avoid bias and show that they enhance the efficiency of Horvitz-Thompson estimators. We derive the efficient influence function for regular, asymptotically linear estimators under Poisson sampling and propose an estimator whose properties hold under complex sampling designs. Our estimator shows rate-double-robustness, ensuring N1/2-normality when nuisance convergence rates exceed N-1/2. The estimator’s asymptotic decomposition requires extending conventional empirical process theory, prompting results in the survey empirical process setting. Our estimator shows superior efficiency in simulations and analysis of the MEPS dataset.

Speaker: Michael Lightfoot

Title: Sparse Functional Principal Component Analysis for Functional Data with Informative Observation Times and Dropout

Authors: Michael Lightfoot, Ana-Maria Staicu 

Abstract: Analysis of sparse functional data has been primarily conducted under the assumption of an uninformative sampling design. We consider the case where the data follow an informative visit process that induces dropout dependent on the functional outcomes. Such a case leads to bias in model component estimation when standard functional data analysis techniques are used. Recent research has presented methods to adjust estimation for informative observation times and censoring, though not while also considering dropout determined by prior functional outcomes. We propose a joint visit process and functional data model following that accounts for both informative observation times and dropout while allowing for sparse, irregular observation times. The performance of the proposed method is shown in numerical studies.