Annals of Applied Statistics最新文献

It is quite common to encounter compositional data in a regression framework in data analysis. When both responses and predictors are compositional, most existing models rely on a family of log-ratio based transformations to move the analysis from the simplex to the reals. This often makes the interpretation of the model more complex. A transformation-free regression model was recently developed, but it only allows for a single compositional predictor. However, many datasets include multiple compositional predictors of interest. Motivated by an application to hydrothermal liquefaction (HTL) data, a novel extension of this transformation-free regression model is provided that allows for two (or more) compositional predictors to be used via a latent variable mixture. A modified expectation-maximization algorithm is proposed to estimate model parameters, which are shown to have natural interpretations. Conformal inference is used to obtain prediction limits on the compositional response. The resulting methodology is applied to the HTL dataset. Extensions to multiple predictors are discussed.

When using electronic health records (EHRs) for clinical and translational research, additional data is often available from external sources to enrich the information extracted from EHRs. For example, academic biobanks have more granular data available, and patient reported data is often collected through small-scale surveys. It is common that the external data is available only for a small subset of patients who have EHR information. We propose efficient and robust methods for building and evaluating models for predicting the risk of binary outcomes using such integrated EHR data. Our method is built upon an idea derived from the two-phase design literature that modeling the availability of a patient's external data as a function of an EHR-based preliminary predictive score leads to effective utilization of the EHR data. Through both theoretical and simulation studies, we show that our method has high efficiency for estimating log-odds ratio parameters, the area under the ROC curve, as well as other measures for quantifying predictive accuracy. We apply our method to develop a model for predicting the short-term mortality risk of oncology patients, where the data was extracted from the University of Pennsylvania hospital system EHR and combined with survey-based patient reported outcome data.

We are motivated by a study that seeks to better understand the dynamic relationship between muscle activation and paw position during locomotion. For each gait cycle in this experiment, activation in the biceps and triceps is measured continuously and in parallel with paw position as a mouse trotted on a treadmill. We propose an innovative general regression method that draws from both ordinary differential equations and functional data analysis to model the relationship between these functional inputs and responses as a dynamical system that evolves over time. Specifically, our model addresses gaps in both literatures and borrows strength across curves estimating ODE parameters across all curves simultaneously rather than separately modeling each functional observation. Our approach compares favorably to related functional data methods in simulations and in cross-validated predictive accuracy of paw position in the gait data. In the analysis of the gait cycles, we find that paw speed and position are dynamically influenced by inputs from the biceps and triceps muscles and that the effect of muscle activation persists beyond the activation itself.

To optimize personalized treatment strategies and extend patients' survival times, it is critical to accurately predict patients' prognoses at all stages, from disease diagnosis to follow-up visits. The longitudinal biomarker measurements during visits are essential for this prediction purpose. Patients' ultimate concerns are cure and survival. However, in many situations, there is no clear biomarker indicator for cure. We propose a comprehensive joint model of longitudinal and survival data and a landmark cure model, incorporating proportions of potentially cured patients. The survival distributions in the joint and landmark models are specified through flexible hazard functions with the proportional hazards as a special case, allowing other patterns such as crossing hazard and survival functions. Formulas are provided for predicting each individual's probabilities of future cure and survival at any time point based on his or her current biomarker history. Simulations show that, with these comprehensive and flexible properties, the proposed cure models outperform standard cure models in terms of predictive performance, measured by the time-dependent area under the curve of receiver operating characteristic, Brier score, and integrated Brier score. The use and advantages of the proposed models are illustrated by their application to a study of patients with chronic myeloid leukemia.

Reconstructing three-dimensional (3D) chromatin structure from conformation capture assays (such as Hi-C) is a critical task in computational biology, since chromatin spatial architecture plays a vital role in numerous cellular processes and direct imaging is challenging. Most existing algorithms that operate on Hi-C contact matrices produce reconstructed 3D configurations in the form of a polygonal chain. However, none of the methods exploit the fact that the target solution is a (smooth) curve in 3D: this contiguity attribute is either ignored or indirectly addressed by imposing spatial constraints that are challenging to formulate. In this paper we develop both B-spline and smoothing spline techniques for directly capturing this potentially complex 1D curve. We subsequently combine these techniques with a Poisson model for contact counts and compare their performance on a real data example. In addition, motivated by the sparsity of Hi-C contact data, especially when obtained from single-cell assays, we appreciably extend the class of distributions used to model contact counts. We build a general distribution-based metric scaling ( $\mathrm{DBMS}$ ) framework from which we develop zero-inflated and Hurdle Poisson models as well as negative binomial applications. Illustrative applications make recourse to bulk Hi-C data from IMR90 cells and single-cell Hi-C data from mouse embryonic stem cells.

Monitoring key elements of disease dynamics (e.g., prevalence, case counts) is of great importance in infectious disease prevention and control, as emphasized during the COVID-19 pandemic. To facilitate this effort, we propose a new capture-recapture (CRC) analysis strategy that adjusts for misclassification stemming from the use of easily administered but imperfect diagnostic test kits, such as rapid antigen test-kits or saliva tests. Our method is based on a recently proposed "anchor stream" design, whereby an existing voluntary surveillance data stream is augmented by a smaller and judiciously drawn random sample. It incorporates manufacturer-specified sensitivity and specificity parameters to account for imperfect diagnostic results in one or both data streams. For inference to accompany case count estimation, we improve upon traditional Wald-type confidence intervals by developing an adapted Bayesian credible interval for the CRC estimator that yields favorable frequentist coverage properties. When feasible, the proposed design and analytic strategy provides a more efficient solution than traditional CRC methods or random sampling-based bias-corrected estimation to monitor disease prevalence while accounting for misclassification. We demonstrate the benefits of this approach through simulation studies and a numerical example that underscore its potential utility in practice for economical disease monitoring among a registered closed population.

We model longitudinal macular thickness measurements to monitor the course of glaucoma and prevent vision loss due to disease progression. The macular thickness varies over a 6 × 6 grid of locations on the retina, with additional variability arising from the imaging process at each visit. currently, ophthalmologists estimate slopes using repeated simple linear regression for each subject and location. To estimate slopes more precisely, we develop a novel Bayesian hierarchical model for multiple subjects with spatially varying population-level and subject-level coefficients, borrowing information over subjects and measurement locations. We augment the model with visit effects to account for observed spatially correlated visit-specific errors. We model spatially varying: (a) intercepts, (b) slopes, and (c) log-residual standard deviations (SD) with multivariate Gaussian process priors with Matérn cross-covariance functions. Each marginal process assumes an exponential kernel with its own SD and spatial correlation matrix. We develop our models for and apply them to data from the Advanced Glaucoma Progression Study. We show that including visit effects in the model reduces error in predicting future thickness measurements and greatly improves model fit.

Causal inference methods can be applied to estimate the effect of a point exposure or treatment on an outcome of interest using data from observational studies. For example, in the Women's Interagency HIV Study, it is of interest to understand the effects of incarceration on the number of sexual partners and the number of cigarettes smoked after incarceration. In settings like this where the outcome is a count, the estimand is often the causal mean ratio, i.e., the ratio of the counterfactual mean count under exposure to the counterfactual mean count under no exposure. This paper considers estimators of the causal mean ratio based on inverse probability of treatment weights, the parametric g-formula, and doubly robust estimation, each of which can account for overdispersion, zero-inflation, and heaping in the measured outcome. Methods are compared in simulations and are applied to data from the Women's Interagency HIV Study.

Case-control experiments are essential to the scientific method, as they allow researchers to test biological hypotheses by looking for differences in outcome between cases and controls. It is then of interest to characterize variation that is enriched in a "foreground" (case) dataset relative to a "background" (control) dataset. For example, in a genomics context, the goal is to identify low-dimensional transcriptional structure unique to patients with certain disease (cases) vs. those without that disease (controls). In this work we propose probabilistic contrastive principal component analysis (PCPCA), a probabilistic dimension reduction method designed for case-control data. We describe inference in PCPCA through a contrastive likelihood and show that our model generalizes PCA, probabilistic PCA, and contrastive PCA. We discuss how to set the tuning parameter in theory and in practice, and we show several of PCPCA's advantages in the analysis of case-control data over related methods, including greater interpretability, uncertainty quantification and principled inference, robustness to noise and missing data, and the ability to generate "foreground-enriched" data from the model. We demonstrate PCPCA's performance on case-control data through a series of simulations, and we successfully identify variation specific to case data in genomic case-control experiments with data modalities, including gene expression, protein expression, and images.

In longitudinal studies, investigators are often interested in understanding how the time since the occurrence of an intermediate event affects a future outcome. The intermediate event is often asymptomatic such that its occurrence is only known to lie in a time interval induced by periodic examinations. We propose a linear regression model that relates the time since the occurrence of the intermediate event to a continuous response at a future time point through a rectified linear unit activation function while formulating the distribution of the time to the occurrence of the intermediate event through the Cox proportional hazards model. We consider nonparametric maximum likelihood estimation with an arbitrary sequence of examination times for each subject. We present an EM algorithm that converges stably for arbitrary datasets. The resulting estimators of regression parameters are consistent, asymptotically normal, and asymptotically efficient. We assess the performance of the proposed methods through extensive simulation studies and provide an application to the Atherosclerosis Risk in Communities Study.