Biometrics最新文献

Investigators often focus on predicting extreme random effects from mixed effects models fitted to longitudinal or clustered data, and on identifying or "flagging" outliers such as poorly performing hospitals or rapidly deteriorating patients. Our recent work with Gaussian outcomes showed that weighted prediction methods can substantially reduce mean square error of prediction for extremes and substantially increase correct flagging rates compared to previous methods, while controlling the incorrect flagging rates. This paper extends the weighted prediction methods to non-Gaussian outcomes such as binary and count data. Closed-form expressions for predicted random effects and probabilities of correct and incorrect flagging are not available for the usual non-Gaussian outcomes, and the computational challenges are substantial. Therefore, our results include the development of theory to support algorithms that tune predictors that we call "self-calibrated" (which control the incorrect flagging rate using very simple flagging rules) and innovative numerical methods to calculate weighted predictors as well as to evaluate their performance. Comprehensive numerical evaluations show that the novel weighted predictors for non-Gaussian outcomes have substantially lower mean square error of prediction at the extremes and considerably higher correct flagging rates than previously proposed methods, while controlling the incorrect flagging rates. We illustrate our new methods using data on emergency room readmissions for children with asthma.

Assessment of multistate disease progression is commonplace in biomedical research, such as in periodontal disease (PD). However, the presence of multistate current status endpoints, where only a single snapshot of each subject's progression through disease states is available at a random inspection time after a known starting state, complicates the inferential framework. In addition, these endpoints can be clustered, and spatially associated, where a group of proximally located teeth (within subjects) may experience similar PD status, compared to those distally located. Motivated by a clinical study recording PD progression, we propose a Bayesian semiparametric accelerated failure time model with an inverse-Wishart proposal for accommodating (spatial) random effects, and flexible errors that follow a Dirichlet process mixture of Gaussians. For clinical interpretability, the systematic component of the event times is modeled using a monotone single index model, with the (unknown) link function estimated via a novel integrated basis expansion and basis coefficients endowed with constrained Gaussian process priors. In addition to establishing parameter identifiability, we present scalable computing via a combination of elliptical slice sampling, fast circulant embedding techniques, and smoothing of hard constraints, leading to straightforward estimation of parameters, and state occupation and transition probabilities. Using synthetic data, we study the finite sample properties of our Bayesian estimates and their performance under model misspecification. We also illustrate our method via application to the real clinical PD dataset.

Daily deaths from an infectious disease provide a means for retrospectively inferring daily incidence, given knowledge of the infection-to-death interval distribution. Existing methods for doing so rely either on fitting simplified non-linear epidemic models to the deaths data or on spline based deconvolution approaches. The former runs the risk of introducing unintended artefacts via the model formulation, while the latter may be viewed as technically obscure, impeding uptake by practitioners. This note proposes a simple simulation based approach to inferring fatal incidence from deaths that requires minimal assumptions, is easy to understand, and allows testing of alternative hypothesized incidence trajectories. The aim is that in any future situation similar to the COVID pandemic, the method can be easily, rapidly, transparently, and uncontroversially deployed as an input to management.

Exposure to environmental pollutants during the gestational period can significantly impact infant health outcomes, such as birth weight and neurological development. Identifying critical windows of susceptibility, which are specific periods during pregnancy when exposure has the most profound effects, is essential for developing targeted interventions. Distributed lag models (DLMs) are widely used in environmental epidemiology to analyze the temporal patterns of exposure and their impact on health outcomes. However, traditional DLMs focus on modeling the conditional mean, which may fail to capture heterogeneity in the relationship between predictors and the outcome. Moreover, when modeling the distribution of health outcomes like gestational birth weight, it is the extreme quantiles that are of most clinical relevance. We introduce 2 new quantile distributed lag model (QDLM) estimators designed to address the limitations of existing methods by leveraging smoothness and shape constraints, such as unimodality and concavity, to enhance interpretability and efficiency. We apply our QDLM estimators to the Colorado birth cohort data, demonstrating their effectiveness in identifying critical windows of susceptibility and informing public health interventions.

We consider a common nonparametric regression setting, where the data consist of a response variable Y, some easily obtainable covariates $mathbf {X}$, and a set of costly covariates $mathbf {Z}$. Before establishing predictive models for Y, a natural question arises: Is it worthwhile to include $mathbf {Z}$ as predictors, given the additional cost of collecting data on $mathbf {Z}$ for both training the models and predicting Y for future individuals? Therefore, we aim to conduct preliminary investigations to infer importance of $mathbf {Z}$ in predicting Y in the presence of $mathbf {X}$. To achieve this goal, we propose a nonparametric variable importance measure for $mathbf {Z}$. It is defined as a parameter that aggregates maximum potential contributions of $mathbf {Z}$ in single or multiple predictive models, with contributions quantified by general loss functions. Considering two-phase data that provide a large number of observations for $(Y,mathbf {X})$ with the expensive $mathbf {Z}$ measured only in a small subsample, we develop a novel approach to infer the proposed importance measure, accommodating missingness of $mathbf {Z}$ in the sample by substituting functions of $(Y,mathbf {X})$ for each individual's contribution to the predictive loss of models involving $mathbf {Z}$. Our approach attains unified and efficient inference regardless of whether $mathbf {Z}$ makes zero or positive contribution to predicting Y, a desirable yet surprising property owing to data incompleteness. As intermediate steps of our theoretical development, we establish novel results in two relevant research areas, semi-supervised inference and two-phase nonparametric estimation. Numerical results from both simulated and real data demonstrate superior performance of our approach.

Directed acyclic graphical models with additive noises are essential in nonlinear causal discovery and have numerous applications in various domains, such as social science and systems biology. Most such models further assume that structural causal functions are additive to ensure causal identifiability and computational feasibility, which may be too restrictive in the presence of causal interactions. Some methods consider general nonlinear causal functions represented by, for example, Gaussian processes and neural networks, to accommodate interactions. However, they are either computationally intensive or lack interpretability. We propose a highly interpretable and computationally feasible approach using trees to incorporate interactions in nonlinear causal discovery, termed tree-based additive noise models. The nature of the tree construction leads to piecewise constant causal functions, making existing causal identifiability results of additive noise models with continuous and smooth causal functions inapplicable. Therefore, we provide new conditions under which the proposed model is identifiable. We develop a recursive algorithm for source node identification and a score-based ordering search algorithm. Through extensive simulations, we demonstrate the utility of the proposed model and algorithms benchmarking against existing additive noise models, especially when there are strong causal interactions. Our method is applied to infer a protein-protein interaction network for breast cancer, where proteins may form protein complexes to perform their functions.

When incorporating historical control data into the analysis of current randomized controlled trial data, it is critical to account for differences between the datasets. When the cause of difference is an unmeasured factor and adjustment for only observed covariates is insufficient, it is desirable to use a dynamic borrowing method that reduces the impact of heterogeneous historical controls. We propose a nonparametric Bayesian approach that addresses between-trial heterogeneity and allows borrowing historical controls homogeneous with the current control. Additionally, to emphasize conflict resolution between historical controls and the current control, we introduce a method based on the dependent Dirichlet process (DP) mixture. The proposed methods can be implemented using the same procedure, regardless of whether the outcome data comprise aggregated study-level data or individual participant data. We also develop a novel index of similarity between the historical and current control data, based on the posterior distribution of the parameter of interest. We conduct a simulation study and analyze clinical trial examples to evaluate the performance of the proposed methods compared to existing methods. The proposed method, based on the dependent DP mixture, can accurately borrow from homogeneous historical controls while reducing the impact of heterogeneous historical controls compared to the typical DP mixture. The proposed methods outperform existing methods in scenarios with heterogeneous historical controls, in which the meta-analytic approach is ineffective.

In recent years, the use of wearable devices, for example, accelerometers, have become increasingly prevalent. Wearable devices enable more accurate real-time tracking of a subject's physical activity (PA) level, such as steps, number of activity bouts, or time in moderate-to-vigorous intensity PA (MVPA), which are important general health markers and can often be represented as counts. These intensive within-subject count data provided by wearable devices, for example, minutes in MVPA summarized per hour across days and even months, allow the possibility for modeling not only the mean PA level, but also the dispersion level for each subject. Especially in the context of daily PA, subjects' dispersion levels are potentially informative in reflecting their exercise patterns: some subjects might exhibit consistent PA across time and can be considered "less dispersed" subjects; while others might have a large amount of PA at a particular time point, while being sedentary for most of the day, and can be considered "more dispersed" subjects. Thus, we propose a negative binomial mixed effects location-scale model to model these intensive longitudinal PA counts and to account for the heterogeneity in both the mean and dispersion level across subjects. Further, to handle the issue of inflated numbers of zeros in the PA data, we also propose a hurdle/zero-inflated version which additionally includes the modeling of the probability of having $>$0 PA levels.

Despite the importance of age-specific fertility for ecology and evolution, the methods for modeling and inference have proven considerably limited. However, other disciplines have long focused on exploring and developing a vast number of models. Here, I provide an overview of the different models proposed since the 1940s by formal demographers, statisticians, and social scientists, most of which are unknown to the ecological and evolutionary communities. I describe how these fall into 2 main categories, namely polynomials and those based on probability density functions. I discuss their merits in terms of their overall behavior and how well they represent the different stages of fertility. Despite many alternative models, inference on age-specific fertility has usually been limited to simple least squares. Although this might be sufficient for human data, I hope to demonstrate that inference requires more sophisticated approaches for ecological and evolutionary datasets. To illustrate how inference and model choice can be achieved on different types of typical ecological and evolutionary data, I present the new R package Bayesian Fertility Trajectory Analysis, which I apply to published aggregated data for lions and baboons. I then conduct a simulation study to test its performance on individual-level data. I show that appropriate inference and model selection can be achieved even when a small number of parents are followed.

A new family of precision Bayesian dose optimization designs, PGen I-II, based on early efficacy, early toxicity, and long-term time to treatment failure is proposed. A PGen I-II design refines a Gen I-II design by accounting for patient heterogeneity characterized by subgroups that may be defined by prognostic levels, disease subtypes, or biomarker categories. The design makes subgroup-specific decisions, which may be to drop an unacceptably toxic or inefficacious dose, randomize patients among acceptable doses, or identify a best dose in terms of treatment success defined in terms of time to failure over long-term follow-up. A piecewise exponential distribution for failure time is assumed, including subgroup-specific effects of dose, response, and toxicity. Latent variables are used to adaptively cluster subgroups found to have similar dose-outcome distributions, with the model simplified to borrow strength between subgroups in the same cluster. Guidelines and user-friendly computer software for implementing the design are provided. A simulation study is reported that shows the PGen I-II design is superior to similarly structured designs that either assume patient homogeneity or conduct separate trials within subgroups.