Statistical Learning and Modeling in Data Analysis: Methods and Applications

Description

The contributions gathered in this book focus on modern methods for statistical learning and modeling in data analysis and present a series of engaging real-world applications. The book covers numerous research topics, ranging from statistical inference and modeling to clustering and factorial methods, from directional data analysis to time series analysis and small area estimation. The applications reflect new analyses in a variety of fields, including medicine, finance, engineering, marketing and cyber risk. The book gathers selected and peer-reviewed contributions presented at the 12th Scientific Meeting of the Classification and Data Analysis Group of the Italian Statistical Society (CLADAG 2019), held in Cassino, Italy, on September 11–13, 2019. CLADAG promotes advanced methodological research in multivariate statistics with a special focus on data analysis and classification, and supports the exchange and dissemination of ideas, methodological concepts, numerical methods, algorithms, and computational and applied results. This book, true to CLADAG’s goals, is intended for researchers and practitioners who are interested in the latest developments and applications in the field of data analysis and classification. Preface Contents Interpreting Effects in Generalized Linear Modeling 1 Introduction 2 Interpreting Effects in Generalized Linear Models for Binary Data 2.1 Alternatives to the Logit and Probit Links with Binary Responses 2.2 Probability Effects Measures for Logistic Models 3 Interpreting Effects in Models for Ordinal Responses 4 Future Potential Research References ACE, AVAS and Robust Data Transformations 1 Introduction 2 Extended Parametric Transformations 3 Robustness and the Fan Plot 4 Augmented Investment Fund Data 5 Robust Analysis 6 Nonparametric Transformations References On Predicting Principal Components Through Linear Mixed Models 1 Introduction 2 Predictive Principal Components Analysis 3 An Application to Some Well-Being Indicators 4 Conclusions and Perspectives References Robust Model-Based Learning to Discover New Wheat Varieties and Discriminate Adulterated Kernels in X-Ray Images 1 Introduction and Motivation 2 RAEDDA Model 2.1 Transductive Learning 2.2 Inductive Learning 3 Anomaly and Novelty Detection in X-Ray Images of Wheat Kernels 4 Conclusions References A Dynamic Model for Ordinal Time Series: An Application to Consumers' Perceptions of Inflation 1 Introduction 2 The Static CUB Model 3 The Dynamic Model for Ordinal Time Series 4 A Case Study: Consumer Inflation Perceptions 5 Final Remarks References Deep Learning to Jointly Analyze Images and Clinical Data for Disease Detection 1 Introduction 2 State of the Art and Challenges of the Medical Deep Learning 3 Material and Methods 3.1 The Data 3.2 Previous Works on the Same Data 3.3 The Model 4 The Results 5 Conclusions References Studying Affiliation Networks Through Cluster CA and Blockmodeling 1 Introduction 2 Factorial Methods and Blockmodeling for Analyzing Affiliation Networks 3 Applying Cluster CA and Blockmodeling to Affiliation Networks 4 A Case Study of Stage Co-productions 4.1 Comparison of Different Ways to Start Initial Cluster Allocation: Randomly and by Blockmodeling Positions 4.2 Cluster CA for Affiliation Networks: Results From Random Start Clustering 5 Concluding Remarks References Sectioning Procedure on Geostatistical Indices Series of Pavement Road Profiles 1 Introduction 2 Geostatistical Tools for Road Pavement Characterization 3 Brief Overview on Road Sectioning Methods 4 Data Collection and Analysis 5 Conclusion References Directional Supervised Learning Through Depth Functions: An Application to ECG Waves Analysis 1 Introduction and Motivations 2 The Arrhythmia Data Set 2.1 Previous Studies 2.2 Scope of the Analysis and Variable Description 3 Directional Depth-Based Supervised Learning Techniques 4 Performance of Depth-Based Classifiers on ECG Waves 5 Final Remarks References Penalized Versus Constrained Approaches for Clusterwise Linear Regression Modeling 1 Introduction 2 The Methodology 2.1 The Constrained Approach 2.2 The Penalized Approach 2.3 Selection of the Tuning Parameter 3 Simulation Study 4 Concluding Remarks References Effect Measures for Group Comparisons in a Two-Component Mixture Model: A Cyber Risk Analysis 1 Introduction 2 cup Models 2.1 Marginal Effect Measures for Covariates in cup Models 2.2 Ordinal Superiority Measures in cup Models 3 Example 4 Discussion and Extension References A Cramér–von Mises Test of Uniformity on the Hypersphere 1 Introduction 2 Background 2.1 Testing Uniformity on Ωq 2.2 Using Projections for Assessing Uniformity 2.3 Projected Uniform Distribution 3 A New Test of Uniformity 3.1 Genesis of the Test Statistic 3.2 U-statistic Form 3.3 Asymptotic Distribution 4 Numerical Experiments 5 Are Venusian Craters Uniformly Distributed? References On Mean And/or Variance Mixtures of Normal Distributions 1 Introduction 2 Mean Mixture of Normal Distributions 2.1 Properties 2.2 Special Cases 3 Mean–Variance Mixture of Normal Distributions 3.1 Properties 3.2 Special Cases 4 Variance Mixture of Normal Distributions 4.1 Properties 4.2 Special Cases 5 Parameter Estimation for MMN, MVMN, and VMN Distributions 6 Conclusions References Robust Depth-Based Inference in Elliptical Models 1 Depth and Illumination in Statistical Analysis 2 Elliptically Symmetric Distributions 2.1 Depth and Illumination of Elliptical Distributions 3 Depth-Based Density Estimation 3.1 The Density Generator as the Volume of Pδ 3.2 The Density Generator as the Probability of Pδ 3.3 Fisher Consistent Estimators of the Density 4 Application References Latent Class Analysis for the Derivation of Marketing Decisions: An Empirical Study for BEV Battery Manufacturers 1 Motivation 2 Latent Class Multinomial Logit Model 3 Empirical Analysis 4 Conclusion References Small Area Estimation Diagnostics: The Case of the Fay–Herriot Model 1 Introduction 2 The Fay–Herriot Model 3 Diagnostics for the Fay–Herriot Model 3.1 The Leverage Matrix 3.2 Influence on the MSE of the EBLUP 3.3 Case Deletion Diagnostics and Cook's Distance 4 An Application to Poverty Data 5 Concluding Remarks References A Comparison Between Methods to Cluster Mixed-Type Data: Gaussian Mixtures Versus Gower Distance 1 Introduction 2 The Model-Based Approach 2.1 Classification, Model Selection, and Identifiability 3 The Gower Distance Method 3.1 k-means 3.2 k-medoids 4 Simulation Study 5 Concluding Remarks References Exploring the Gender Gap in Erasmus Student Mobility Flows 1 Introduction 2 Data and Methods 3 Results References