Advances and Innovations in Statistics and Data Science

Description

This book highlights selected papers from the 4th ICSA-Canada Chapter Symposium, as well as invited articles from established researchers in the areas of statistics and data science. It covers a variety of topics, including methodology development in data science, such as methodology in the analysis of high dimensional data, feature screening in ultra-high dimensional data and natural language ranking; statistical analysis challenges in sampling, multivariate survival models and contaminated data, as well as applications of statistical methods. With this book, readers can make use of frontier research methods to tackle their problems in research, education, training and consultation. Preface Part I: Methodology Development in Data Science (Chapters 1–6) Part II: Challenges in Statistical Analysis (Chapters 7–15) Contents Contributors Part I Methodology Development in Data Science MiRNA–Gene Activity Interaction Networks (miGAIN): Integrated Joint Models of miRNA–Gene Targeting and Disturbance in Signaling Pathways 1 Introduction 2 Data 3 Methods 4 Results 4.1 Simulations 4.2 Data Analysis 5 Discussion References Robust Feature Screening for Ultrahigh-Dimensional Censored Data Subject to Measurement Error 1 Introduction 2 Notation and Framework 2.1 Survival Data 2.2 Distance Correlation of Two Random Variables 2.3 Feature Screening for Censored Data with Precise Measurements 3 Feature Screening for Censored Data with Error-Prone Covariates 3.1 Measurement Error Model 3.2 Feature Screening with Measurement Error Effects Accommodated 3.3 Asymptotic Results 4 Iteration Algorithm 5 Numerical Studies 5.1 Simulation Setup 5.2 Simulation Results 5.3 Analysis of Mantle Cell Lymphoma Microarray Data 6 Discussion Appendix A. Technical Lemmas B. Proofs of Main Theorems B.1 Proof of Theorem 1 B.2 Proof of Theorem 2 References Simultaneous Control of False Discovery Rate and Sensitivity Using Least Angle Regressions in High-Dimensional Data Analysis 1 Introduction 2 Methodology 2.1 Cosin Distribution 2.2 Selection Criteria 3 Numerical Studies 3.1 Simulation Studies 3.1.1 Scenario I: Compound Symmetry Structure of 3.1.2 Scenario II: Auto-Regressive Correlation 3.2 A Real-Data Application 4 Discussion References Minimum Wasserstein Distance Estimator Under Finite Location-Scale Mixtures 1 Introduction 2 Wasserstein Distance and the Minimum Distance Estimator 2.1 Wasserstein Distance 2.2 Minimum Wasserstein Distance Estimator 2.3 Consistency of MWDE 2.4 Numerical Solution to MWDE 2.5 Penalized Maximum Likelihood Estimator 3 Experiments 3.1 Performance Measure 3.2 Performance Under Homogeneous Model 3.3 Efficiency and Robustness Under Finite Location-Scale Mixtures 3.3.1 Efficiency 3.3.2 Robustness 3.4 Image Segmentation 4 Conclusion Appendix Numerically Friendly Expression of W2(FN, F(·|G)) References An Entropy-Based Comment Ranking Method with Word Embedding Clustering 1 Introduction 1.1 How to Judge a Comment's Quality? 1.2 Ranking Comments Using Entropy 1.3 Text Representation 2 Bag-of-Words Model with Word Embedding Clusters 3 Ranking Comments with General Entropy 4 Experiment with Amazon Review Data References A Robust Approach to Statistical Quality Control for High-Dimensional Non-Normal Data 1 Introduction 2 Notations and Preliminaries 3 Modification and Robustness 3.1 Model and Assumptions 3.2 Statistic, Its Limit, and Robustness 4 Simulations 5 Discussion 6 Basic Moments 7 Proof of Lemma 1 8 Proof of Theorem 1 9 Proof of Theorem 2 References Part II Challenges in Statistical Analysis Functional Linear Regression for Partially Observed Functional Data 1 Introduction 2 Functional Linear Model 3 Estimation Methods 3.1 Partially Observed Functional Data Without Measurement Error 3.2 Partially Observed Functional Data with Measurement Error 4 Simulation Studies 5 Real Data Analysis 6 Discussion Appendix References Profile Estimation of Generalized Semiparametric Varying-Coefficient Additive Models for Longitudinal Data with Within-Subject Correlations 1 Introduction 2 Profile GEE Estimation Procedures 2.1 Model Estimation Using Fixed Working Covariance Matrices 2.2 Estimation of the Within-Subject Covariance Matrix 2.2.1 Estimation of Marginal Variance 2.2.2 Estimation of Correlation Coefficients 2.3 Profile Estimation via Quadratic Inference Function 2.4 Computational Algorithms 2.5 Choices of Kernel Function, Bandwidth, and Link Function 3 Simulation Studies 3.1 Continuous Longitudinal Responses 3.2 Discrete Longitudinal Responses 4 Concluding Remarks References Sieve Estimation of Semiparametric Linear Transformation Model with Left-Truncated and Current Status Data 1 Introduction 2 Sieve Maximum Likelihood Estimation 3 Efficient Estimation 4 Asymptotic Properties 5 Simulation Study 6 Real Data Analysis 7 Concluding Remarks and Future Work Appendix References A Review of Flexible Transformations for Modeling Compositional Data 1 Introduction 2 Transformations for Compositional Data 2.1 Isometric Log-Ratio Transformation 2.2 α-transformation 2.3 α-Folding Transformation 3 Real-life Data Applications 4 Conclusion References Identifiability and Estimation of Autoregressive ARCH Models with Measurement Error 1 Introduction 2 The Model and Identifiability 3 GMM Estimation 4 Testing for Measurement Error 5 Impact of Measurement Error 6 Finite Sample Properties 7 Conclusions and Discussion Appendix Regularity Assumptions and Mathematical Proofs Regularity Assumptions Proof of Theorem 1 Proof of Theorem 2 References Modal Regression for Skewed, Truncated, or Contaminated Data with Outliers 1 Introduction 2 Linear Modal Regression 2.1 Introduction of Linear Modal Regression 2.2 Asymptotic Properties 2.3 Estimation Algorithm 2.4 Prediction Intervals Based on Modal Regression 3 Nonparametric Modal Regression 3.1 Estimating Uni-Modal Regression 3.2 Estimating Multi-Modal Regression 4 Semiparametric Modal Regression 5 Discussion Appendix References Spatial Multilevel Modelling in the Galveston Bay Recovery Study Survey 1 Introduction 2 Sampling Design 3 Survey Weights 4 Inferring Inclusion Probabilities from the Weights 5 Spatial Multilevel Model 6 A Bayesian Analysis 7 Frequentist Composite Likelihood Analysis 7.1 Estimating Function System for Mean Parameters 7.2 Decomposition of the Error Term 7.3 Estimating Equation System for Variance Components 7.4 Point Estimation 7.5 Uncertainty Estimation 8 Discussion and Conclusions References Efficient Experimental Design for Lasso Regression 1 Introduction 2 Methodology 2.1 A Construction Method Using Orthogonal Arrays 2.2 A Construction Method Using the Kronecker Product 3 Numerical Illustration 4 Discussion References A Selective Overview of Statistical Methods for Identification of the Treatment-Sensitive Subsets of Patients 1 Introduction 2 Statistical Methods for Treatment-Sensitive Subset Identification Based on Survival Times 2.1 An Approach Based on a Biomarker-Adaptive Threshold Design 2.2 A Hierarchical Bayesian Method 2.3 A Procedure Based on a Single-index Threshold Cox Model 2.4 An Interaction Tree Approach 3 Statistical Methods for Treatment-Sensitive Subset Identification Based on Longitudinal Measurements 3.1 A Procedure Based on Multilevel Models 3.2 A Prediction Model Approach 3.3 A Procedure Based on a Threshold Linear Mixed Model 4 Discussions and Future Work References Index