Associate Professor of Biostatistics
Yang Feng is an Associate Professor of Biostatistics. He received his B.S. in mathematics from the University of Science and Technology of China and his Ph.D. in Operations Research from Princeton University.
Dr. Feng's research interests include machine learning with applications to public health, high-dimensional statistics, network models, nonparametric statistics, and bioinformatics. He has published in The Annals of Statistics, Journal of the American Statistical Association, Journal of the Royal Statistical Society Series B, Journal of Machine Learning Research, International Journal of Epidemiology, and Science Advances. Feng serves on the editorial boards of the Journal of Business & Economic Statistics, Statistica Sinica, Stat, and Statistical Analysis and Data Mining: The ASA Data Science Journal.
Prior to joining NYU, Feng was an Associate Professor of Statistics and an affiliated member in the Data Science Institute at Columbia University. He is an elected member of the International Statistical Institute and a recipient of the NSF CAREER award.
B.S. in Mathematics, University of Science and Technology of China, Hefei, ChinaPh.D. in Operations Research, Princeton University, Princeton, NJ
BioinformaticsBiostatisticsHigh-dimensional data analysis/integrationMachine learningModeling Social and Behavioral DynamicsNonparametric statistics
RaSE: A Variable Screening Framework via Random Subspace EnsemblesTian, Y., & Feng, Y. (n.d.).
Journal titleJournal of the American Statistical Association
Page(s)457-468AbstractVariable screening methods have been shown to be effective in dimension reduction under the ultra-high dimensional setting. Most existing screening methods are designed to rank the predictors according to their individual contributions to the response. As a result, variables that are marginally independent but jointly dependent with the response could be missed. In this work, we propose a new framework for variable screening, random subspace ensemble (RaSE), which works by evaluating the quality of random subspaces that may cover multiple predictors. This new screening framework can be naturally combined with any subspace evaluation criterion, which leads to an array of screening methods. The framework is capable to identify signals with no marginal effect or with high-order interaction effects. It is shown to enjoy the sure screening property and rank consistency. We also develop an iterative version of RaSE screening with theoretical support. Extensive simulation studies and real-data analysis show the effectiveness of the new screening framework.
Spectral Clustering via Adaptive Layer Aggregation for Multi-Layer NetworksHuang, S., Weng, H., & Feng, Y. (n.d.).
Journal titleJournal of Computational and Graphical Statistics
Page(s)1170-1184AbstractOne of the fundamental problems in network analysis is detecting community structure in multi-layer networks, of which each layer represents one type of edge information among the nodes. We propose integrative spectral clustering approaches based on effective convex layer aggregations. Our aggregation methods are strongly motivated by a delicate asymptotic analysis of the spectral embedding of weighted adjacency matrices and the downstream k-means clustering, in a challenging regime where community detection consistency is impossible. In fact, the methods are shown to estimate the optimal convex aggregation, which minimizes the misclustering error under some specialized multi-layer network models. Our analysis further suggests that clustering using Gaussian mixture models is generally superior to the commonly used k-means in spectral clustering. Extensive numerical studies demonstrate that our adaptive aggregation techniques, together with Gaussian mixture model clustering, make the new spectral clustering remarkably competitive compared to several popularly used methods. Supplementary materials for this article are available online.
Variable selection for high-dimensional generalized linear model with block-missing dataHe, Y., Feng, Y., & Song, X. (n.d.).
Journal titleScandinavian Journal of Statistics
Page(s)1279-1297AbstractIn modern scientific research, multiblock missing data emerges with synthesizing information across multiple studies. However, existing imputation methods for handling block-wise missing data either focus on the single-block missing pattern or heavily rely on the model structure. In this study, we propose a single regression-based imputation algorithm for multiblock missing data. First, we conduct a sparse precision matrix estimation based on the structure of block-wise missing data. Second, we impute the missing blocks with their means conditional on the observed blocks. Theoretical results about variable selection and estimation consistency are established in the context of a generalized linear model. Moreover, simulation studies show that compared with existing methods, the proposed imputation procedure is robust to various missing mechanisms because of the good properties of regression imputation. An application to Alzheimer's Disease Neuroimaging Initiative data also confirms the superiority of our proposed method.
A likelihood-ratio type test for stochastic block models with bounded degreesYuan, M., Feng, Y., & Shang, Z. (n.d.).
Journal titleJournal of Statistical Planning and Inference
Page(s)98-119AbstractA fundamental problem in network data analysis is to test Erdös–Rényi model [Formula presented] versus a bisection stochastic block model [Formula presented], where a,b>0 are constants that represent the expected degrees of the graphs and n denotes the number of nodes. This problem serves as the foundation of many other problems such as testing-based methods for determining the number of communities (Bickel and Sarkar, 2016; Lei, 2016) and community detection (Montanari and Sen, 2016). Existing work has been focusing on growing-degree regime a,b→∞ (Bickel and Sarkar, 2016; Lei, 2016; Montanari and Sen, 2016; Banerjee and Ma, 2017; Banerjee, 2018; Gao and Lafferty, 2017a,b) while leaving the bounded-degree regime untreated. In this paper, we propose a likelihood-ratio (LR) type procedure based on regularization to test stochastic block models with bounded degrees. We derive the limit distributions as power Poisson laws under both null and alternative hypotheses, based on which the limit power of the test is carefully analyzed. We also examine a Monte-Carlo method that partly resolves the computational cost issue. The proposed procedures are examined by both simulated and real-world data. The proof depends on a contiguity theory developed by Janson (1995).
Association of hyperglycemia and molecular subclass on survival in IDH-wildtype glioblastomaLiu, E. K., Vasudevaraja, V., Sviderskiy, V. O., Feng, Y., Tran, I., Serrano, J., Cordova, C., Kurz, S. C., Golfinos, J. G., Sulman, E. P., Orringer, D. A., Placantonakis, D., Possemato, R., & Snuderl, M. (n.d.).
Journal titleNeuro-Oncology Advances
Issue1AbstractBackground: Hyperglycemia has been associated with worse survival in glioblastoma. Attempts to lower glucose yielded mixed responses which could be due to molecularly distinct GBM subclasses. Methods: Clinical, laboratory, and molecular data on 89 IDH-wt GBMs profiled by clinical next-generation sequencing and treated with Stupp protocol were reviewed. IDH-wt GBMs were sub-classified into RTK I (Proneural), RTK II (Classical) and Mesenchymal subtypes using whole-genome DNA methylation. Average glucose was calculated by time-weighting glucose measurements between diagnosis and last follow-up. Results: Patients were stratified into three groups using average glucose: tertile one (<100 mg/dL), tertile two (100-115 mg/dL), and tertile three (>115 mg/dL). Comparison across glucose tertiles revealed no differences in performance status (KPS), dexamethasone dose, MGMT methylation, or methylation subclass. Overall survival (OS) was not affected by methylation subclass (P =. 9) but decreased with higher glucose (P =. 015). Higher glucose tertiles were associated with poorer OS among RTK I (P =. 08) and mesenchymal tumors (P =. 05), but not RTK II (P =. 99). After controlling for age, KPS, dexamethasone, and MGMT status, glucose remained significantly associated with OS (aHR = 5.2, P =. 02). Methylation clustering did not identify unique signatures associated with high or low glucose levels. Metabolomic analysis of 23 tumors showed minimal variation across metabolites without differences between molecular subclasses. Conclusion: Higher average glucose values were associated with poorer OS in RTKI and Mesenchymal IDH-wt GBM, but not RTKII. There were no discernible epigenetic or metabolomic differences between tumors in different glucose environments, suggesting a potential survival benefit to lowering systemic glucose in selected molecular subtypes.
Clinical, Pathological, and Molecular Characteristics of Diffuse Spinal Cord GliomasGarcia, M. R., Feng, Y., Vasudevaraja, V., Galbraith, K., Serrano, J., Thomas, C., Radmanesh, A., Hidalgo, E. T., Harter, D. H., Allen, J. C., Gardner, S. L., Osorio, D. S., William, C. M., Zagzag, D., Boué, D. R., & Snuderl, M. (n.d.).
Journal titleJournal of Neuropathology and Experimental Neurology
Page(s)865-872AbstractDiffuse spinal cord gliomas (SCGs) are rare tumors associated with a high morbidity and mortality that affect both pediatric and adult populations. In this retrospective study, we sought to characterize the clinical, pathological, and molecular features of diffuse SCG in 22 patients with histological and molecular analyses. The median age of our cohort was 23.64 years (range 1-82) and the overall median survival was 397 days. K27M mutation was significantly more prevalent in males compared to females. Gross total resection and chemotherapy were associated with improved survival, compared to biopsy and no chemotherapy. While there was no association between tumor grade, K27M status (p = 0.366) or radiation (p = 0.772), and survival, males showed a trend toward shorter survival. K27M mutant tumors showed increased chromosomal instability and a distinct DNA methylation signature.
Community detection with nodal information: Likelihood and its variational approximationWeng, H., & Feng, Y. (n.d.).
Issue1AbstractCommunity detection is one of the fundamental problems in the study of network data. Most existing community detection approaches only consider edge information as inputs, and the output could be suboptimal when nodal information is available. In such cases, it is desirable to leverage nodal information for the improvement of community detection accuracy. Towards this goal, we propose a flexible network model incorporating nodal information and develop likelihood-based inference methods. For the proposed methods, we establish favorable asymptotic properties as well as efficient algorithms for computation. Numerical experiments show the effectiveness of our methods in utilizing nodal information across a variety of simulated and real network data sets.
Large-scale model selection in misspecified generalized linear modelsDemirkaya, E., Feng, Y., Basu, P., & Lv, J. (n.d.).
Page(s)123-136AbstractModel selection is crucial both to high-dimensional learning and to inference for contemporary big data applications in pinpointing the best set of covariates among a sequence of candidate interpretable models. Most existing work implicitly assumes that the models are correctly specified or have fixed dimensionality, yet both model misspecification and high dimensionality are prevalent in practice. In this paper, we exploit the framework of model selection principles under the misspecified generalized linear models presented in Lv Liu (2014), and investigate the asymptotic expansion of the posterior model probability in the setting of high-dimensional misspecified models. With a natural choice of prior probabilities that encourages interpretability and incorporates the Kullback-Leibler divergence, we suggest using the high-dimensional generalized Bayesian information criterion with prior probability for large-scale model selection with misspecification. Our new information criterion characterizes the impacts of both model misspecification and high dimensionality on model selection. We further establish the consistency of covariance contrast matrix estimation and the model selection consistency of the new information criterion in ultrahigh dimensions under some mild regularity conditions. Our numerical studies demonstrate that the proposed method enjoys improved model selection consistency over its main competitors.
Model Averaging for Nonlinear Regression ModelsFeng, Y., Liu, Q., Yao, Q., & Zhao, G. (n.d.).
Journal titleJournal of Business and Economic Statistics
Page(s)785-798AbstractThis article considers the problem of model averaging for regression models that can be nonlinear in their parameters and variables. We consider a nonlinear model averaging (NMA) framework and propose a weight-choosing criterion, the nonlinear information criterion (NIC). We show that up to a constant, NIC is an asymptotically unbiased estimator of the risk function under nonlinear settings with some mild assumptions. We also prove the optimality of NIC and show the convergence of the model averaging weights. Monte Carlo experiments reveal that NMA leads to relatively lower risks compared with alternative model selection and model averaging methods in most situations. Finally, we apply the NMA method to predicting the individual wage, where our approach leads to the lowest prediction errors in most cases.
Targeting Predictors Via Partial Distance Correlation With Applications to Financial ForecastingYousuf, K., & Feng, Y. (n.d.).
Journal titleJournal of Business and Economic Statistics
Page(s)1007-1019AbstractHigh-dimensional time series datasets are becoming increasingly common in various fields of economics and finance. Given the ubiquity of time series data, it is crucial to develop efficient variable screening methods that use the unique features of time series. This article introduces several model-free screening methods based on partial distance correlation and developed specifically to deal with time-dependent data. Methods are developed both for univariate models, such as nonlinear autoregressive models with exogenous predictors (NARX), and multivariate models such as linear or nonlinear VAR models. Sure screening properties are proved for our methods, which depend on the moment conditions, and the strength of dependence in the response and covariate processes, amongst other factors. We show the effectiveness of our methods via extensive simulation studies and an application on forecasting U.S. market returns.
TESTING COMMUNITY STRUCTURE FOR HYPERGRAPHSYuan, M., Liu, R., Feng, Y., & Shang, Z. (n.d.).
Journal titleAnnals of Statistics
Page(s)147-169AbstractMany complex networks in the real world can be formulated as hypergraphs where community detection has been widely used. However, the fundamental question of whether communities exist or not in an observed hypergraph remains unclear. This work aims to tackle this important problem. Specifically, we systematically study when a hypergraph with community structure can be successfully distinguished from its Erdos-Rényi counterpart, and propose concrete test statistics when the models are distinguishable. The main contribution of this paper is threefold. First, we discover a phase transition in the hyperedge probability for distinguishability. Second, in the bounded-degree regime, we derive a sharp signal-to-noise ratio (SNR) threshold for distinguishability in the special two-community 3- uniform hypergraphs, and derive nearly tight SNR thresholds in the general two-community m-uniform hypergraphs. Third, in the dense regime, we propose a computationally feasible test based on sub-hypergraph counts, obtain its asymptotic distribution, and analyze its power. Our results are further extended to nonuniform hypergraphs in which a new test involving both edge and hyperedge information is proposed. The proofs rely on Janson's contiguity theory (Combin. Probab. Comput. 4 (1995) 369-405), a high-moments driven asymptotic normality result by Gao andWormald (Probab. Theory Related Fields 130 (2004) 368-376), and a truncation technique for analyzing the likelihood ratio.
Transfer Learning Under High-Dimensional Generalized Linear ModelsTian, Y., & Feng, Y. (n.d.).
Journal titleJournal of the American Statistical AssociationAbstractIn this work, we study the transfer learning problem under high-dimensional generalized linear models (GLMs), which aim to improve the fit on target data by borrowing information from useful source data. Given which sources to transfer, we propose a transfer learning algorithm on GLM, and derive its (Formula presented.) -estimation error bounds as well as a bound for a prediction error measure. The theoretical analysis shows that when the target and sources are sufficiently close to each other, these bounds could be improved over those of the classical penalized estimator using only target data under mild conditions. When we don’t know which sources to transfer, an algorithm-free transferable source detection approach is introduced to detect informative sources. The detection consistency is proved under the high-dimensional GLM transfer learning setting. We also propose an algorithm to construct confidence intervals of each coefficient component, and the corresponding theories are provided. Extensive simulations and a real-data experiment verify the effectiveness of our algorithms. We implement the proposed GLM transfer learning algorithms in a new R package glmtrans, which is available on CRAN. Supplementary materials for this article are available online.
Analytical performance of lateral flow immunoassay for SARS-CoV-2 exposure screening on venous and capillary blood samplesBlack, M. A., Shen, G., Feng, X., Garcia Beltran, W. F., Feng, Y., Vasudevaraja, V., Allison, D., Lin, L. H., Gindin, T., Astudillo, M., Yang, D., Murali, M., Iafrate, A. J., Jour, G., Cotzia, P., & Snuderl, M. (n.d.).
Journal titleJournal of Immunological Methods
Volume489AbstractObjectives: We validate the use of a lateral flow immunoassay (LFI) intended for rapid screening and qualitative detection of anti-SARS-CoV-2 IgM and IgG in serum, plasma, and whole blood, and compare results with ELISA. We also seek to establish the value of LFI testing on blood obtained from a capillary blood sample. Methods: Samples collected by venous blood draw and finger stick were obtained from patients with SARS-CoV-2 detected by RT-qPCR and control patients. Samples were tested with Biolidics 2019-nCoV IgG/IgM Detection Kit lateral flow immunoassay, and antibody calls were compared with ELISA. Results: Biolidics LFI showed clinical sensitivity of 92% with venous blood at 7 days after PCR diagnosis of SARS-CoV-2. Test specificity was 92% for IgM and 100% for IgG. There was no significant difference in detecting IgM and IgG with Biolidics LFI and ELISA at D0 and D7 (p = 1.00), except for detection of IgM at D7 (p = 0.04). Capillary blood of SARS-CoV-2 patients showed 93% sensitivity for antibody detection. Conclusions: Clinical performance of Biolidics 2019-nCoV IgG/IgM Detection Kit is comparable to ELISA and was consistent across sample types. This provides an opportunity for decentralized rapid testing and may allow point-of-care and longitudinal self-testing for the presence of anti-SARS-CoV-2 antibodies.
Association of body composition parameters measured on CT with risk of hospitalization in patients with Covid-19Chandarana, H., Pisuchpen, N., Krieger, R., Dane, B., Mikheev, A., Feng, Y., Kambadakone, A., & Rusinek, H. (n.d.).
Journal titleEuropean Journal of Radiology
Volume145AbstractPurpose: To assess prognostic value of body composition parameters measured at CT to predict risk of hospitalization in patients with COVID-19 infection. Methods: 177 patients with SARS-CoV-2 infection and with abdominopelvic CT were included in this retrospective IRB approved two-institution study. Patients were stratified based on disease severity as outpatients (no hospital admission) and patients who were hospitalized (inpatients). Two readers blinded to the clinical outcome segmented axial CT images at the L3 vertebral body level for visceral adipose tissue (VAT), subcutaneous adipose tissue (SAT), muscle adipose tissue (MAT), muscle mass (MM). VAT to total adipose tissue ratio (VAT/TAT), MAT/MM ratio, and muscle index (MI) at L3 were computed. These measures, along with detailed clinical risk factors, were compared in patients stratified by severity. Various logistic regression clinical and clinical + imaging models were compared to discriminate inpatients from outpatients. Results: There were 76 outpatients (43%) and 101 inpatients. Male gender (p = 0.013), age (p = 0.0003), hypertension (p = 0.0003), diabetes (p = 0.0001), history of cardiac disease (p = 0.007), VAT/TAT (p < 0.0001), and MAT/MM (p < 0.0001), but not BMI, were associated with hospitalization. A clinical model (age, gender, BMI) had AUC of 0.70. Addition of VAT/TAT to the clinical model improved the AUC to 0.73. Optimal model that included gender, BMI, race (Black), MI, VAT/TAT, as well as interaction between gender and VAT/TAT and gender and MAT/MM demonstrated the highest AUC of 0.83. Conclusion: MAT/MM and VAT/TAT provides important prognostic information in predicting patients with COVID-19 who are likely to require hospitalization.
Comparison of solid tissue sequencing and liquid biopsy accuracy in identification of clinically relevant gene mutations and rearrangements in lung adenocarcinomasLin, L. H., Allison, D. H., Feng, Y., Jour, G., Park, K., Zhou, F., Moreira, A. L., Shen, G., Feng, X., Sabari, J., Velcheti, V., Snuderl, M., & Cotzia, P. (n.d.).
Journal titleModern Pathology
Page(s)2168-2174AbstractScreening for therapeutic targets is standard of care in the management of advanced non-small cell lung cancer. However, most molecular assays utilize tumor tissue, which may not always be available. “Liquid biopsies” are plasma-based next generation sequencing (NGS) assays that use circulating tumor DNA to identify relevant targets. To compare the sensitivity, specificity, and accuracy of a plasma-based NGS assay to solid-tumor-based NGS we retrospectively analyzed sequencing results of 100 sequential patients with lung adenocarcinoma at our institution who had received concurrent testing with both a solid-tissue-based NGS assay and a commercially available plasma-based NGS assay. Patients represented both new diagnoses (79%) and disease progression on treatment (21%); the majority (83%) had stage IV disease. Tissue-NGS identified 74 clinically relevant mutations, including 52 therapeutic targets, a sensitivity of 94.8%, while plasma-NGS identified 41 clinically relevant mutations, a sensitivity of 52.6% (p < 0.001). Tissue-NGS showed significantly higher sensitivity and accuracy across multiple patient subgroups, both in newly diagnosed and treated patients, as well as in metastatic and nonmetastatic disease. Discrepant cases involved hotspot mutations and actionable fusions including those in EGFR, ALK, and NTRK1. In summary, tissue-NGS detects significantly more clinically relevant alterations and therapeutic targets compared to plasma-NGS, suggesting that tissue-NGS should be the preferred method for molecular testing of lung adenocarcinoma when tissue is available. Plasma-NGS can still play an important role when tissue testing is not possible. However, given its low sensitivity, a negative result should be confirmed with a tissue-based assay.
Imbalanced classification: A paradigm-based reviewFeng, Y., Zhou, M., & Tong, X. (n.d.).
Journal titleStatistical Analysis and Data Mining
Page(s)383-406AbstractA common issue for classification in scientific research and industry is the existence of imbalanced classes. When sample sizes of different classes are imbalanced in training data, naively implementing a classification method often leads to unsatisfactory prediction results on test data. Multiple resampling techniques have been proposed to address the class imbalance issues. Yet, there is no general guidance on when to use each technique. In this article, we provide a paradigm-based review of the common resampling techniques for binary classification under imbalanced class sizes. The paradigms we consider include the classical paradigm that minimizes the overall classification error, the cost-sensitive learning paradigm that minimizes a cost-adjusted weighted type I and type II errors, and the Neyman–Pearson paradigm that minimizes the type II error subject to a type I error constraint. Under each paradigm, we investigate the combination of the resampling techniques and a few state-of-the-art classification methods. For each pair of resampling techniques and classification methods, we use simulation studies and a real dataset on credit card fraud to study the performance under different evaluation metrics. From these extensive numerical experiments, we demonstrate under each classification paradigm, the complex dynamics among resampling techniques, base classification methods, evaluation metrics, and imbalance ratios. We also summarize a few takeaway messages regarding the choices of resampling techniques and base classification methods, which could be helpful for practitioners.
Mediation effect selection in high-dimensional and compositional microbiome dataZhang, H., Chen, J., Feng, Y., Wang, C., Li, H., & Liu, L. (n.d.).
Journal titleStatistics in Medicine
Page(s)885-896AbstractThe microbiome plays an important role in human health by mediating the path from environmental exposures to health outcomes. The relative abundances of the high-dimensional microbiome data have an unit-sum restriction, rendering standard statistical methods in the Euclidean space invalid. To address this problem, we use the isometric log-ratio transformations of the relative abundances as the mediator variables. To select significant mediators, we consider a closed testing-based selection procedure with desirable confidence. Simulations are provided to verify the effectiveness of our method. As an illustrative example, we apply the proposed method to study the mediation effects of murine gut microbiome between subtherapeutic antibiotic treatment and body weight gain, and identify Coprobacillus and Adlercreutzia as two significant mediators.
RaSE: Random subspace ensemble classificationTian, Y., & Feng, Y. (n.d.).
Journal titleJournal of Machine Learning Research
Volume22AbstractWe propose a exible ensemble classification framework, Random Subspace Ensemble (RaSE), for sparse classification. In the RaSE algorithm, we aggregate many weak learners, where each weak learner is a base classifier trained in a subspace optimally selected from a collection of random subspaces. To conduct subspace selection, we propose a new criterion, ratio information criterion (RIC), based on weighted Kullback-Leibler divergence. The theoretical analysis includes the risk and Monte-Carlo variance of the RaSE classifier, establishing the screening consistency and weak consistency of RIC, and providing an upper bound for the misclassification rate of the RaSE classifier. In addition, we show that in a high-dimensional framework, the number of random subspaces needs to be very large to guarantee that a subspace covering signals is selected. Therefore, we propose an iterative version of the RaSE algorithm and prove that under some specific conditions, a smaller number of generated random subspaces are needed to find a desirable subspace through iteration. An array of simulations under various models and real-data applications demonstrate the effectiveness and robustness of the RaSE classifier and its iterative version in terms of low misclassification rate and accurate feature ranking. The RaSE algorithm is implemented in the R package RaSEn on CRAN.
The Interplay of Demographic Variables and Social Distancing Scores in Deep Prediction of U.S. COVID-19 CasesTang, F., Feng, Y., Chiheb, H., & Fan, J. (n.d.).
Journal titleJournal of the American Statistical Association
Page(s)492-506AbstractWith the severity of the COVID-19 outbreak, we characterize the nature of the growth trajectories of counties in the United States using a novel combination of spectral clustering and the correlation matrix. As the United States and the rest of the world are still suffering from the effects of the virus, the importance of assigning growth membership to counties and understanding the determinants of the growth is increasingly evident. For the two communities (faster versus slower growth trajectories) we cluster the counties into, the average between-group correlation is 88.4% whereas the average within-group correlations are 95.0% and 93.8%. The average growth rate for one group is 0.1589 and 0.1704 for the other, further suggesting that our methodology captures meaningful differences between the nature of the growth across various counties. Subsequently, we select the demographic features that are most statistically significant in distinguishing the communities: number of grocery stores, number of bars, Asian population, White population, median household income, number of people with the bachelor’s degrees, and population density. Lastly, we effectively predict the future growth of a given county with a long short-term memory (LSTM) recurrent neural network using three social distancing scores. The best-performing model achieves a median out-of-sample R 2 of 0.6251 for a four-day ahead prediction and we find that the number of communities and social distancing features play an important role in producing a more accurate forecasting. This comprehensive study captures the nature of the counties’ growth in cases at a very micro-level using growth communities, demographic factors, and social distancing performance to help government agencies utilize known information to make appropriate decisions regarding which potential counties to target resources and funding to. Supplementary materials for this article, including a standardized description of the materials available for reproducing the work, are available as an online supplement.
Visceral adipose tissue in patients with COVID-19: risk stratification for severityChandarana, H., Dane, B., Mikheev, A., Taffel, M. T., Feng, Y., & Rusinek, H. (n.d.).
Journal titleAbdominal Radiology
Page(s)818-825AbstractPURPOSE: To assess visceral (VAT), subcutaneous (SAT), and total adipose tissue (TAT) estimates at abdominopelvic CT in COVID-19 patients with different severity, and analyze Body Mass Index (BMI) and CT estimates of fat content in patients requiring hospitalization.METHODS: In this retrospective IRB approved HIPPA compliant study, 51 patients with SARS-CoV-2 infection with abdominopelvic CT were included. Patients were stratified based on disease severity as outpatient (no hospital admission) and patients who were hospitalized. Subset of hospitalized patient required mechanical ventilation (MV). A radiologist blinded to the clinical outcome evaluated single axial slice on CT at L3 vertebral body for VAT L3, SAT L3, TAT L3, and VAT/TAT L3. These measures along with age, gender, and BMI were compared. A clinical model that included age, sex, and BMI was compared to clinical + CT model that also included VAT L3 to discriminate hospitalized patients from outpatients. RESULTS: There were ten outpatients and 41 hospitalized patients. 11 hospitalized patients required MV. There were no significant differences in age and BMI between the hospitalized and outpatients (all p > 0.05). There was significantly higher VAT L3 and VAT/TAT L3 in hospitalized patients compared to the outpatients (all p < 0.05). Area under the curve (AUC) of the clinical + CT model was higher compared to the clinical model (AUC 0.847 versus 0.750) for identifying patients requiring hospitalization. CONCLUSION: Higher VAT L3 was observed in COVID-19 patients that required hospitalization compared to the outpatients, and addition of VAT L3 to the clinical model improved AUC in discriminating hospitalized from outpatients in this preliminary study.
A projection-based conditional dependence measure with applications to high-dimensional undirected graphical modelsFan, J., Feng, Y., & Xia, L. (n.d.).
Journal titleJournal of Econometrics
Page(s)119-139AbstractMeasuring conditional dependence is an important topic in econometrics with broad applications including graphical models. Under a factor model setting, a new conditional dependence measure based on projection is proposed. The corresponding conditional independence test is developed with the asymptotic null distribution unveiled where the number of factors could be high-dimensional. It is also shown that the new test has control over the asymptotic type I error and can be calculated efficiently. A generic method for building dependency graphs without Gaussian assumption using the new test is elaborated. We show the superiority of the new method, implemented in the R package pgraph, through simulation and real data studies.
Accounting for incomplete testing in the estimation of epidemic parametersBetensky, R. A., & Feng, Y. (n.d.).
Journal titleInternational Journal of Epidemiology
Nested model averaging on solution path for high-dimensional linear regressionFeng, Y., & Liu, Q. (n.d.).
Issue1AbstractWe study the nested model averaging method on the solution path for a high-dimensional linear regression problem. In particular, we propose to combine model averaging with regularized estimators (e.g., lasso, elastic net, and Sorted L-One Penalized Estimation [SLOPE]) on the solution path for high-dimensional linear regression. In simulation studies, we first conduct a systematic investigation on the impact of predictor ordering on the behaviour of nested model averaging, and then show that nested model averaging with lasso, elastic net and SLOPE compares favourably with other competing methods, including the infeasible lasso, elastic, net and SLOPE with the tuning parameter optimally selected. A real data analysis on predicting the per capita violent crime in the United States shows outstanding performance of the nested model averaging with lasso.
Neyman-pearson classification: Parametrics and sample size requirementTong, X., Xia, L., Wang, J., & Feng, Y. (n.d.).
Journal titleJournal of Machine Learning Research
Volume21AbstractThe Neyman-Pearson (NP) paradigm in binary classification seeks classifiers that achieve a minimal type II error while enforcing the prioritized type I error controlled under some user-specified level α. This paradigm serves naturally in applications such as severe disease diagnosis and spam detection, where people have clear priorities among the two error types. Recently, Tong et al. (2018) proposed a nonparametric umbrella algorithm that adapts all scoring-type classification methods (e.g., logistic regression, support vector machines, random forest) to respect the given type I error (i.e., conditional probability of classifying a class 0 observation as class 1 under the 0-1 coding) upper bound α with high probability, without specific distributional assumptions on the features and the responses. Universal the umbrella algorithm is, it demands an explicit minimum sample size requirement on class 0, which is often the more scarce class, such as in rare disease diagnosis applications. In this work, we employ the parametric linear discriminant analysis (LDA) model and propose a new parametric thresholding algorithm, which does not need the minimum sample size requirements on class 0 observations and thus is suitable for small sample applications such as rare disease diagnosis. Leveraging both the existing nonparametric and the newly proposed parametric thresholding rules, we propose four LDA-based NP classifiers, for both low- and high-dimensional settings. On the theoretical front, we prove NP oracle inequalities for one proposed classifier, where the rate for excess type II error benefits from the explicit parametric model assumption. Furthermore, as NP classifiers involve a sample splitting step of class 0 observations, we construct a new adaptive sample splitting scheme that can be applied universally to NP classifiers, and this adaptive strategy reduces the type II error of these classifiers. The proposed NP classifiers are implemented in the R package nproc.
On the estimation of correlation in a binary sequence modelWeng, H., & Feng, Y. (n.d.).
Journal titleJournal of Statistical Planning and Inference
Page(s)123-137AbstractWe consider a binary sequence generated by thresholding a hidden continuous sequence. The hidden variables are assumed to have a compound symmetry covariance structure with a single parameter characterizing the common correlation. We study the parameter estimation problem under such one-parameter models. We demonstrate that maximizing the likelihood function does not yield consistent estimates for the correlation. We then formally prove the nonestimability of the parameter by deriving a non-vanishing minimax lower bound. This counter-intuitive phenomenon provides an interesting insight that one-bit information of each latent variable is not sufficient to consistently recover their common correlation. On the other hand, we further show that trinary data generated from the hidden variables can consistently estimate the correlation with parametric convergence rate. Thus we reveal a phase transition phenomenon regarding the discretization of latent continuous variables while preserving the estimability of the correlation. Numerical experiments are performed to validate the conclusions.