Biometrical Journal最新文献

We study time-dependent mediators in survival analysis using a treatment separation approach due to Didelez [Lifetime Data Analysis 25, no. 4: 593–610] and based on earlier work by Robins and Richardson [Causality and Psychopathology: Finding the Determinants of Disorders and Their Cures, 103–158. Oxford University Press]. This approach avoids nested counterfactuals and cross-world assumptions which are otherwise common in mediation analysis. The causal model of treatment, mediators, covariates, confounders, and outcome is represented by directed acyclic graphs (DAGs). However, the DAGs tend to be very complex when we have measurements at many time points. We therefore suggest using so-called rolled graphs in which a node represents an entire coordinate process instead of a single random variable, leading us to far simpler graphical representations. The rolled graphs are not necessarily acyclic; they can be analyzed by $�\delta$-separation which is the appropriate graphical separation criterion in this class of graphs and analogous to $�d$-separation. In particular, $�\delta$-separation is a graphical tool for evaluating if the conditions of the mediation analysis are met, or if unmeasured confounders influence the estimated effects. We also state a mediational g-formula. This is similar to the approach in Vansteelandt et al. [Statistics in Medicine 38, no. 24: 4828–4840], although that paper has a different conceptual basis. Finally, we apply this framework to a statistical model based on a Cox model with an added treatment effect.

This paper is concerned with estimation and testing for treatment effects with multivariate outcomes. It primarily focuses on the situation where imperfect diagnostic tools are used to classify subjects into different groups. Oftentimes, there are more expensive and/or invasive diagnostic tools to accurately determine the subjects' status or conditions, yielding partially validated data on a smaller number of subjects. We propose moment-based approaches for estimating and testing treatment effects. We compare our methods with maximum likelihood approach using the EM algorithm, which requires strong assumptions and bears computational burden, and with traditional methods, which ignore the diagnostic tool's imperfection. The proposed methods show advantages in terms of coverage probability, computations efficiency, and robustness. The application of the methods is illustrated with gene-expression data from the Genes-environments & Admixture in Latino Americans (GALA) II study of asthma in Hispanic/Latino children.

High-dimensional data often arise from clinical genomics research to infer relevant predictors of a particular trait. A way to improve the predictive performance is by incorporating information about the predictors obtained from existing from prior knowledge or previous studies. Such information is also referred to as “co-data.” To this aim, we develop a novel Bayesian model for including co-data in a high-dimensional regression framework, termed informative Horseshoe regression (infHS). The proposed approach regresses the prior variances of the regression parameters on the co-data variables, improving variable selection and prediction. We implement both a Gibbs sampler and a Variational approximation algorithm. The former is suited for applications of moderate dimensions which, besides prediction, target posterior inference, whereas the latter's computational efficiency allows handling a very large number of variables. We show the benefits of including co-data through a simulation study. Lastly, we demonstrate that infHS outperforms competing approaches in two genomics applications.

Modeling events associated with discrete-valued observations arises in several practical situations. Until now, research on statistical methods for discrete data has primarily focused on modeling count data. Nevertheless, discrete observations that may assume any value in the set of integers $��Z�=�\{�\text{…}�,�-�2�,�-�1�,�0�,�1�,�2�,�\text{…}�\}�$ are also found in various contexts. This paper introduces a general parametric modeling framework for the analysis of integer-valued data, with applications to paired discrete observations. The proposed model is based on the modified skew discrete Laplace distribution. Our approach enables a straightforward interpretation of regression coefficients in terms of mean and dispersion, while properly accounting for the discrete nature of the data. We adopt a frequentist approach to perform inference and define diagnostic tools to assess goodness-of-fit. Additionally, we conduct several simulation studies to examine the asymptotic properties of the estimators and test statistics, as well as the distribution of the residuals. We illustrate the usefulness of the proposed model with two real datasets: one from an experimental study conducted in a French penitentiary, and another involving diagnostic imaging to assess kidney function through dynamic and static scintigraphy. Estimation and inference procedures for the new regression model are implemented in the R package sdlrm.

Prediction model selection and assessment are primary objectives of many statistical analyses. Covariance-based penalty estimators provide analytic estimates of the optimism associated with naive training error estimates for multiple classes of prediction models and error assessment rules. While the majority of work on covariance-based penalties has focused on prediction for uncensored data, little attention has been given to time-to-event data. In this article, we consider estimating the optimism for survival prediction models assessed via the Brier score. We first analytically derive an expression of the optimism in a single group scenario with uncensored data based on a reformulation of the optimism. With the same reformulation, we propose an algorithm to estimate the optimism for Cox's proportional hazards regression under a general prediction setting involving covariates and right censoring. We verify the derived theory and demonstrate the applicability of the proposed algorithm via simulation studies. Finally, we illustrate the utility of our new covariance-based penalty estimator through an application predicting time to hemodialysis access failure among patients with end-stage renal disease using data from the United States Renal Data System.

Functional data analysis has received significant attention due to its frequent occurrence in modern applications, such as in the medical field, where electrocardiograms or electroencephalograms can be used for a better understanding of various medical conditions. Due to the infinite-dimensional nature of functional elements, the current work focuses on dimension reduction techniques. This study shifts its focus to modeling the conditional quantiles of functional data, noting that existing works are limited to quantitative predictors. Consequently, we introduce the first approach to partial dimension reduction for the conditional quantiles under the presence of both functional and categorical predictors. We present the proposed algorithm and derive the convergence rates of the estimators. Moreover, we demonstrate the finite sample performance of the method using simulation examples and a real dataset based on functional magnetic resonance imaging.

In the competing risks setting, the $�t$-year absolute risk for a specific time $�t$ (e.g., 2 years), also called the cumulative incidence function at time $�t$, is often interesting to estimate. It is routinely estimated using the nonparametric Aalen–Johansen estimator. This estimator handles right-censored data and has desirable large sample properties, as it is the nonparametric maximum likelihood estimator (NPMLE). Inference for comparing absolute risks, via either a risk difference or a risk ratio, can therefore be done via usual asymptotic normal approximations and the delta method. However, the small sample performances of this approach are not fully satisfactory. Especially, (i) coverage of confidence intervals may be inaccurate and (ii) comparisons made using a risk ratio and a risk difference can lead to inconsistent conclusions, in terms of statistical significance. We, therefore, introduce an alternative empirical likelihood approach. One advantage of this approach is that it always leads to consistent conclusions when comparing absolute risks via a risk ratio and a risk difference, in terms of significance. Simulation results also suggest that small sample inference using this approach can be more accurate. We present the computation of confidence intervals and p-values using this approach and the asymptotic properties that justify them. We provide formulas and algorithms to compute constrained NPMLE, from which empirical likelihood ratios and inference procedures are derived. The novel approach has been implemented in the timeEL package for R, and some of its advantages are demonstrated via reproducible analyses of bone marrow transplant data.