Search

In a one-way analysis-of-variance (ANOVA) model, the number of pairwise comparisons can become large even with a moderate number of groups. Motivated by this, we consider a regime with a growing number of groups and prove that, when testing pairwise comparisons, the Benjamini-Hochberg (BH) procedure can asymptotically control false discoveries, despite the fact that the involved t-statistics do not exhibit the well-known positive dependence structure required for exact false discovery rate (FDR) control. Following Tukey's perspective that the difference between the means of any two groups cannot be exactly zero, our main result provides control over the directional false discovery rate and directional false discovery proportion. A key technical contribution of our work is demonstrating that the dependence among the t-statistics is sufficiently weak to establish the convergence result typically required for asymptotic FDR control. Our analysis does not rely on conventional assumptions such as normality, variance homogeneity, or a balanced design, thereby offering a theoretical foundation for applications in more general settings. [ABSTRACT FROM AUTHOR]

Co-inertia analysis (CIA) is a multivariate analysis method that assesses relationships and trends in two sets of data. It has been effectively employed in the integrative analysis of high-dimensional multi-omics datasets. Recently, penalized CIA methods have been introduced to enhance the interpretability by inducing sparsity in the loading vectors. However, challenges persist in ensuring that non-zero elements in the estimated vector genuinely represent significant features. To address these challenges, we propose a penalized CIA method that controls the false discovery rate (FDR) using sorted l-1 penalized estimation (SLOPE). This approach allows for simultaneous FDR control and sparsity induction in the estimated vectors. Extensive simulation studies demonstrate the performance compared to the existing CIA method. Additionally, we apply our methods to the integrative analysis of NCI60 data to show its effectiveness in real-world scenarios. [ABSTRACT FROM AUTHOR]

ABSTRACT Currently in wheat breeding, genome wide association studies (GWAS) have successfully revealed the genetic basis of complex traits such as nitrogen use efficiency (NUE) and its biological processes. In the GWAS model, thresholding is common strategy to indicate deviation of expected range of p‐value(s), and it can be used to find the distribution of true positive associations under or over of test statistics. Therefore, the threshold plays a critical role to identify reliable and significant associations in wide genome, while the proportion of false positive results is relatively low. The problem of multiple comparisons arises when a statistical analysis involves multiple simultaneous statistical tests, each of them has the potential to be a discovery. There are several ways to address this problem, including the family‐wise error rate and false discovery rate (FDR), raw and adjusted p‐value(s), consideration of threshold coherence and consonance, and the properties of proportional hypothesis tests in the threshold definition. We encountered some limitations in the definition of FDR threshold, particularly in the upper bounds of linear and nonlinear approaches. We emphasize that empirical null distributions based on permutation test can be useful when the assumption of linear or parametric FDR approaches do not hold. Nevertheless, we believe that it is necessary to utilize modern statistical optimization techniques to evaluate the stability and performance of our results and to select significant FDR threshold. By incorporating the neural network algorithm, it is possible to improve the reliability of FDR threshold and increase the probability of identifying true genetic associations while minimizing the risk of false positives in GWAS results.

Background: Accurate type 1 diabetes prediction is important to facilitate screening for pre-clinical type 1 diabetes to enable potential early disease-modifying interventions and to reduce the risk of severe presentation with diabetic ketoacidosis. We aimed to assess the generalisability of a prediction model developed in children followed from birth. Additionally, we sought to create an application for easy calculation and visualization of individualized risk prediction., Methods: We developed and refined a stratified prediction model combining a genetic risk score, age, islet autoantibodies, and family history using data from children followed since birth by The Environmental Determinants of Diabetes in the Young (TEDDY) study. We tested the validity of the model through external validation in the Type 1 Diabetes TrialNet Pathway to Prevention study, which conducts cross-sectional screening in relatives of people with type 1 diabetes. We recalibrated the model by adjusting for baseline risk and selection criteria in TrialNet using logistic recalibration to improve calibration across all ages., Results: The study included 7,798 TEDDY and 4,068 TrialNet participants, with 305 (4%) and 1,373 (34%) developing type 1 diabetes, respectively. The combined model showed similar discriminative ability in autoantibody-positive individuals across TEDDY and TrialNet (p=0.14), but inferior calibration in TrialNet (Brier score 0.40 [0.38,0.43]). Adjustment for baseline risk and selection criteria in TrialNet using logistic recalibration improved calibration across all ages (Brier score 0.16 [0.14,0.17]; p<0.001). A web calculator was developed to visualise individual risk estimates (https://t1dpredictor.diabetesgenes.org)., Conclusions: A stratified model of type 1 diabetes genetic risk score, family history, age, and autoantibody status accurately predicts type 1 diabetes risk, but may need recalibration according to screening stategy., Competing Interests: Competing interests: The authors have no other relevant conflicts of interest to disclose.