Stats最新文献

Imputing missing values in high-dimensional time-series data remains a significant challenge in statistics and machine learning. Although various methods have been proposed in recent years, many struggle with limitations and reduced accuracy, particularly when the missing rate is high. In this work, we present a novel f-divergence-based bidirectional generative adversarial imputation network, tf-BiGAIN, designed to address these challenges in time-series data imputation. Unlike traditional imputation methods, tf-BiGAIN employs a generative model to synthesize missing values without relying on distributional assumptions. The imputation process is achieved by training two neural networks, implemented using bidirectional modified gated recurrent units, with f-divergence serving as the objective function to guide optimization. Compared to existing deep learning-based methods, tf-BiGAIN introduces two key innovations. First, the use of f-divergence provides a flexible and adaptable framework for optimizing the model across diverse imputation tasks, enhancing its versatility. Second, the use of bidirectional gated recurrent units allows the model to leverage both forward and backward temporal information. This bidirectional approach enables the model to effectively capture dependencies from both past and future observations, enhancing its imputation accuracy and robustness. We applied tf-BiGAIN to analyze two real-world time-series datasets, demonstrating its superior performance in imputing missing values and outperforming existing methods in terms of accuracy and robustness.

Meta-analysis aggregates information across related studies to provide more reliable statistical inference and has been a vital tool for assessing the safety and efficacy of many high-profile pharmaceutical products. A key challenge in conducting a meta-analysis is that the number of related studies is typically small. Applying classical methods that are asymptotic in the number of studies can compromise the validity of inference, particularly when heterogeneity across studies is present. Moreover, serious adverse events are often rare and can result in one or more studies with no events in at least one study arm. Practitioners remove studies in which no events have occurred in one or both arms or apply arbitrary continuity corrections (e.g., adding one event to arms with zero events) to stabilize or define effect estimates in such settings, which can further invalidate subsequent inference. To address these significant practical issues, we introduce an exact inference method for random effects meta-analysis of a treatment effect in the two-sample setting with rare events, which we coin "XRRmeta". In contrast to existing methods, XRRmeta provides valid inference for meta-analysis in the presence of between-study heterogeneity and when the event rates, number of studies, and/or the within-study sample sizes are small. Extensive numerical studies indicate that XRRmeta does not yield overly conservative inference. We apply our proposed method to two real-data examples using our open-source R package.

Background: Previous research has identified differences in e-cigarette use and socioeconomic factors between different racial groups However, there is little research examining specific risk factors contributing to the racial differences.

Objective: This study sought to identify racial disparities in e-cigarette use and to determine risk factors that help explain these differences.

Methods: We used Wave 5 (2018-2019) of the Adult Population Assessment of Tobacco and Health (PATH) Study. First, we conducted descriptive statistics of e-smoking across our risk factor variables. Next, we used multiple logistic regression to check the risk effects by adjusting all covariates. Finally, we conducted a mediation analysis to determine whether identified factors showed evidence of influencing the association between race and e-cigarette use. All analyses were performed in R or SAS. The R package mma was used for the mediation analysis.

Results: Between Hispanic and non-Hispanic White populations, our potential risk factors collectively explain 17.5% of the racial difference, former cigarette smoking explains 7.6%, receiving e-cigarette advertising 2.6%, and perception of e-cigarette harm explains 27.8% of the racial difference. Between non-Hispanic Black and non-Hispanic White populations, former cigarette smoking, receiving e-cigarette advertising, and perception of e-cigarette harm explain 5.2%, 1.8%, and 6.8% of the racial difference, respectively. E-cigarette use is most prevalent in the non-Hispanic White population compared to non-Hispanic Black and Hispanic populations, which may be explained by former cigarette smoking, exposure to e-cigarette advertising, and e-cigarette harm perception.

Conclusions: These findings suggest that racial differences in e-cigarette use may be reduced by increasing knowledge of the dangers associated with e-cigarette use and reducing exposure to e-cigarette advertisements. This comprehensive analysis of risk factors can be used to significantly guide smoking cessation efforts and address potential health burden disparities arising from differences in e-cigarette usage.

We investigate a semiparametric generalized partially linear regression model that accommodates missing outcomes, with some covariates modeled parametrically and others nonparametrically. We propose a class of augmented inverse probability weighted (AIPW) kernel-profile estimating equations. The nonparametric component is estimated using AIPW kernel estimating equations, while parametric regression coefficients are estimated using AIPW profile estimating equations. We demonstrate the doubly robust nature of the AIPW estimators for both nonparametric and parametric components. Specifically, these estimators remain consistent if either the assumed model for the probability of missing data or that for the conditional mean of the outcome, given covariates and auxiliary variables, is correctly specified, though not necessarily both simultaneously. Additionally, the AIPW profile estimator for parametric regression coefficients is consistent and asymptotically normal under the semiparametric model defined by the generalized partially linear model on complete data, assuming that the missing data mechanism is missing at random. When both working models are correctly specified, this estimator achieves semiparametric efficiency, with its asymptotic variance reaching the efficiency bound. We validate our approach through simulations to assess the finite sample performance of the proposed estimators and apply the method to a study that investigates risk factors associated with myocardial ischemia.

Prostate cancer (PCa) poses a significant challenge because of the difficulty in identifying aggressive tumors, leading to overtreatment and missed personalized therapies. Although only 8% of cases progress beyond the prostate, the accurate prediction of aggressiveness remains crucial. Thus, this study focused on studying retinoblastoma phosphorylated at Serine 249 (Phospho-Rb S249), N-cadherin, β-catenin, and E-cadherin as biomarkers for identifying aggressive PCa using a logistic regression model and a classification and regression tree (CART). Using immunohistochemistry (IHC), we targeted the expression of these biomarkers in PCa tissues and correlated their expression with clinicopathological data of the tumor. The results showed a negative correlation between E-cadherin and β-catenin with aggressive tumor behavior, whereas Phospho-Rb S249 and N-cadherin positively correlated with increased tumor aggressiveness. Furthermore, patients were stratified based on Gleason scores and E-cadherin staining patterns to evaluate their capability for early identification of aggressive PCa. Our findings suggest that the classification tree is the most effective method for measuring the utility of these biomarkers in clinical practice, incorporating β-catenin, tumor grade, and Gleason grade as relevant determinants for identifying patients with Gleason scores ≥ 4 + 3. This study could potentially benefit patients with aggressive PCa by enabling early disease detection and closer monitoring.

People who inject drugs (PWID) have an increased risk of HIV infection partly due to injection behaviors often related to opioid use. Medications for opioid use disorder (MOUD) have been shown to reduce HIV infection risk, possibly by reducing injection risk behaviors. MOUD may benefit individuals who do not receive it themselves but are connected through social, sexual, or drug use networks with individuals who are treated. This is known as spillover. Valid estimation of spillover in network studies requires considering the network's community structure. Communities are groups of densely connected individuals with sparse connections to other groups. We analyzed a network of 277 PWID and their contacts from the Transmission Reduction Intervention Project. We assessed the effect of MOUD on reductions in injection risk behaviors and the possible benefit for network contacts of participants treated with MOUD. We identified communities using modularity-based methods and employed inverse probability weighting with community-level propensity scores to adjust for measured confounding. We found that MOUD may have beneficial spillover effects on reducing injection risk behaviors. The magnitudes of estimated effects were sensitive to the community detection method. Careful consideration should be paid to the significance of community structure in network studies evaluating spillover.