Data Analysis and Classification (eBook)

Preface 5
List of Referees 8
Contents 9
Contributors 15
Part I Key-note 21
Clustering of High-Dimensional and Correlated Data 22
1 Introduction 22
2 Definition of Mixture Models 23
3 Maximum Likelihood Estimation 24
4 Choice of Starting Values for the EM Algorithm 24
5 Clustering via Normal Mixtures 25
6 Factor Analysis Model for Dimension Reduction 26
7 Some Recent Extensions for High-Dimensional Data 27
8 Mixtures of Normal Components with Random Effects 28
References 30
Statistical Methods for Cryptography 31
1 Introduction 31
1.1 Different disciplines in cryptography 32
2 Prime Numbers 33
2.1 Tests of primality 33
2.2 Deterministic tests 34
2.3 Some deterministic tests 35
3 The Sum Modulo m of Statistical Variables 36
References 39
Part II Cluster Analysis 40
An Algorithm for Earthquakes Clustering Based on MaximumLikelihood 41
1 Introduction 41
2 Conditional Intensity Function of the Clustering Procedure 42
2.1 The ETAS Model 43
2.2 Intensity Function for a Particular ClusteredInhomogeneous Poisson Process 44
3 The Proposed Clustering Method 45
3.1 Finding a Candidate Cluster and Likelihood Changes 45
3.2 The Algorithm of Clustering 46
4 Application to a Real Catalog and Final Remarks 47
References 48
A Two-Step Iterative Procedure for Clustering of Binary Sequences 49
1 Introduction 49
2 Clustering and Dimensionality Reduction 50
3 Example 53
References 56
Clustering Linear Models Using Wasserstein Distance 57
1 Introduction 57
2 Input Data and the Clustering Algorithm 58
3 Dynamic Clustering of Linear Models 59
3.1 Wasserstein Distance for Distributions 60
3.2 Representation and Allocation Functions 61
4 An Application Using Bank of Italy Household Survey Data 62
5 Conclusions and Perspectives 64
References 64
Comparing Approaches for Clustering Mixed Mode Data: An Application in Marketing Research 65
1 Introduction 65
2 Obtaining Partitions with Mixed Mode Data 66
3 Some Illustrative Applications 67
4 Discussion 72
References 72
The Progressive Single Linkage Algorithm Based on Minkowski Ultrametrics 74
1 Introduction 74
2 Ultrametric Approximations 75
3 The Progressive Single Linkage Algorithm 76
4 Some Applications to Real Data 77
5 Conclusions 79
Appendix 80
References 81
Visualization of Model-Based Clustering Structures 82
1 Introduction 82
2 Dimension Reduction for Model-Based Clustering 83
3 Visualization of Clustering Structures 84
4 Examples 86
4.1 Overlapping Clusters with Unconstrained Covariances 86
4.2 High Dimensional Mixture of Two Normal Distributions 87
4.3 Wisconsin Diagnostic Breast Cancer Data 87
5 Comments and Extensions 89
References 89
Part III Multidimensional Scaling 91
Models for Asymmetry in Proximity Data 92
1 Introduction 92
2 A Class of Scalar Product Models 92
3 The Analysis of Skew-Symmetry with External Information 94
4 Conclusions 96
References 97
Intimate Femicide in Italy: A Model to Classify How KillingsHappened 98
1 Introduction 98
2 National and International Scenario 99
3 Data and Method 101
4 The Main Results 102
References 104
Two-Dimensional Centrality of Asymmetric Social Network 105
1 Introduction 105
2 The Procedure 106
3 The Data 107
4 The Analysis 107
5 Result 108
6 Discussion 109
References 111
The Forward Search for Classical Multidimensional Scaling When the Starting Data Matrix Is Known 113
1 Introduction 113
2 Classical Multidimensional Scaling and the Forward Search 114
3 A Case Study: Linosa Dataset 116
4 Conclusions 120
References 121
Part IV Multivariate Analysis and Application 122
Discriminant Analysis on Mixed Predictors 123
1 Introduction 123
2 Mixed Discriminant Predictors 124
3 Application Example 126
3.1 Predictor Analysis 126
3.2 Discriminant Analysis 128
3.3 Comparison 128
4 Conclusion 130
References 130
A Statistical Calibration Model for Affymetrix Probe Level Data 131
1 Introduction 131
2 A Brief Review on preprocessing Methods 132
3 The Proposed Calibration Method 133
4 A Comparison with the Most Popular Methods 135
5 Conclusions 137
References 138
A Proposal to Fuzzify Categorical Variables in Operational Risk Management 139
1 Fuzzy Approach 139
2 The Problem 140
3 The Proposal 141
4 Results 143
5 Conclusions 144
References 145
Common Optimal Scaling for Customer Satisfaction Models: A Point to Cobb–Douglas' Form 146
1 Features of a Customer Satisfaction model 146
2 Categorical Regression with Common Optimal Scaling 147
2.1 The Pattern of the Model 148
2.1.1 The Algorithm of the Parameters Estimation 149
3 Multiplicative Models for CS 149
3.1 Some Observations 150
4 A Theory About Overall CS 151
5 Conclusions 152
References 152
Structural Neural Networks for Modeling Customer Satisfaction 154
1 Introduction 154
2 Customer Satisfaction and PLS Path Modeling 155
3 Customer Satisfaction and Neural Networks 156
4 A Structural Neural Network for Modeling CS 157
5 Concluding Remarks 161
References 161
Dimensionality of Scores Obtained witha Paired-Comparison Tournament Systemof Questionnaire Items 163
1 Preference Data Collection 163
2 Preference Data 164
3 Scoring Algorithms 165
4 Dimensions in Preference Data 166
5 Conclusions 170
References 170
Using Rasch Measurement to Assess the Roleof the Traditional Family in Italy 171
1 Introduction 171
2 Data and Descriptive Analysis 172
3 Method 173
4 Discussion 175
References 177
Preserving the Clustering Structure by a Projection PursuitApproach 178
1 Introduction 178
2 Projection Pursuit for Preserving the Clustering Structure 180
2.1 The Critical Bandwidth to Test Multimodality 180
2.2 Projection Pursuit Using the Adjusted Critical Bandwidth 182
3 Numerical Results 182
3.1 A Simulation Study 182
3.2 Real Data Applications 183
4 Concluding Remarks 185
References 185
Association Rule Mining of Multimedia Content 186
1 Introduction 186
2 Syntactic Analysis of Video Data 187
3 Semantic Analysis 189
4 Association Rules 190
5 Synergy Effects 191
References 192
Part V Classification and Classification Tree 194
Automatic Dictionary- and Rule-Based Systems for Extracting Information from Text 195
1 Introduction 195
2 A Model for Creating a Meta-dictionary by Means of a Hybrid System 197
3 Application to the Istat TUS Survey 199
References 204
Several Computational Studies About Variable Selection for Probabilistic Bayesian Classifiers 205
1 Introduction 205
2 Bayesian Networks and Classification 206
2.1 Learning Bayesian Networks 207
2.2 Classifiers Based on Bayesian Networks 207
3 Feature Subset Selection for Classification 207
4 Experimental Results and Conclusions 208
References 212
Semantic Classification and Co-occurrences: A Methodfor the Rules Production for the Information Extractionfrom Textual Data 214
1 Introduction 214
2 The Analyzed Corpus and Semantic Classification 215
3 Rules Production Using Co-occurrences and Collocations 215
4 Future Developments and Improvements 220
References 221
The Effectiveness of University Education: A Structural Equation Model 222
1 Introduction 222
2 The Model 223
3 Main Results 226
4 Conclusions 228
References 229
Simultaneous Threshold Interaction Detection in BinaryClassification 230
1 Introduction 230
2 Modeling Interaction Effects in Regression Analysis 231
3 The Trunk Model 233
4 Empirical Evidence 234
5 Concluding Remarks 236
References 237
Detecting Subset of Classifiers for Multi-attribute ResponsePrediction 238
1 Introduction 238
2 The SASSC Algorithm 239
3 Analyzing the Letter Recognition Dataset 242
4 Concluding Remarks 244
References 245
Clustering Textual Data by Latent Dirichlet Allocation: Applications and Extensions to Hierarchical Data 246
1 Introduction 246
2 The LDA Model 247
3 Considering Documents Structure: An Extension of the LDA Model 249
4 Generalizing the Prior Enrichment: Collapsed Variational Bayes 251
5 Discussion 253
References 253
Multilevel Latent Class Models for Evaluation of Long-term Care Facilities 254
1 Introduction and Data 254
2 The Model 255
3 The Results 257
4 Conclusions 261
References 261
Author–Coauthor Social Networks and Emerging ScientificSubfields 262
1 Introduction 262
2 Distribution of Tie Strength 264
3 Distribution of Clique Size 266
4 Random Graph Model for Preferential Attachment 267
5 The Emergence of Scientific Subfields 268
6 The Network of Well-Established Scholars 270
7 Conclusions 272
References 273
Part VI Statistical Models 274
A Hierarchical Model for Time Dependent Multivariate Longitudinal Data 275
1 Introduction 275
2 Model-Based Approach to Three-Way Data Clustering 276
3 Multivariate Hidden Markov Model for Three-Way Data Clustering 277
4 Computational Details 279
5 Simulation Results 280
6 Conclusion 282
References 283
Covariate Error Bias Effects in Dynamic Regression Model Estimation and Improvement in the Prediction by Covariate Local Clusters 284
1 Introduction 284
2 Bias Effects in Dynamic Regression Models with Errors in the Covariate 285
3 A Local Cluster Kalman Filter 286
4 Simulation Experiments 287
5 An Application 289
6 Conclusions 291
References 291
Local Multilevel Modeling for Comparisons of InstitutionalPerformance 292
1 Introduction 292
2 Capturing Local Behaviour 293
2.1 Mixture Modeling 293
2.2 Cluster-Weighted Modeling 294
3 The Proposal: Local Multilevel Modeling 295
4 An Example 296
5 Conclusions and Further Research 298
References 299
Modelling Network Data: An Introduction to Exponential Random Graph Models 300
1 Introduction 300
2 Exponential Random Graph Models for Social Networks 301
3 The Collaboration Network of Italian Scholars on Population Studies 303
4 Model Estimation Results 305
5 Concluding Remarks 307
References 307
Part VII Latent Variables 309
An Analysis of Earthquakes Clustering Based on a Second-Order Diagnostic Approach 310
1 Introduction 310
2 Second-Order Residual Analysis 311
2.1 The Weighted Process and Its Second-Order Properties 311
2.1.1 The Weighted Spectrum 312
2.1.2 The Weighted Correlation Integral 312
3 Space–Time ETAS Model 313
4 Nonparametric Estimation and Diagnostics 314
5 Conclusion 317
References 317
Latent Regression in Rasch Framework 319
1 Introduction 319
2 Longitudinal Latent Regression Model 320
3 Latent Regression Rasch Model with Missing Data 322
4 Conclusion 325
References 326
A Multilevel Latent Variable Model for Multidimensional Longitudinal Data 328
1 Introduction 328
2 Model for Continuous Responses 329
2.1 Model Specification 329
2.2 Estimation 331
2.3 Application to a Real Data Set 333
3 Conclusion 335
References 335
Turning Point Detection Using Markov Switching Models with Latent Information 336
1 Introduction 336
2 The Generalized Hamilton Model with Latent Information 337
3 Identifying and Forecasting USA and Japanese Turning Points 339
References 343
Part VIII Knowledge Extraction from Temporal Data 344
Statistical and Numerical Algorithms for Time SeriesClassification 345
1 Introduction 345
2 A Simulation Experiment 348
3 Application to Real Data 350
4 Concluding Remarks 351
References 351
Mining Time Series Data: A Selective Survey 353
1 Introduction 353
2 Comparing Time Series Shape 354
3 Criteria Based on Fourier and Wavelet Analysis 355
4 Structural Dissimilarity 357
5 Final Remarks 358
References 359
Predictive Dynamic Models for SMEs 361
1 Introduction 361
2 Default Estimation: A Methodological Proposal 362
3 Data Sources 364
4 Application 364
5 Conclusion 365
References 366
Clustering Algorithms for Large Temporal Data Sets 367
1 The Framework: Temporal Data Mining 367
2 Temporal Cluster Analysis 368
3 Clustering Algorithms: Applicability to Large Temporal Datasets 370
4 An Example of Application on a Radar Satellite Data Base 372
5 Concluding Remarks 373
References 374
Part IX Outlier Detection and Robust Methods 376
Robust Clustering for Performance Evaluation 377
1 Introduction 377
2 Mahalanobis Distances and the Forward Search 378
3 Example 379
References 385
Outliers Detection Strategy for a Curve Clustering Algorithm 387
1 Introduction 387
2 Dynamical Curves Clustering with Free knots Spline Estimation 388
3 An Improvement of DCC& FSE Algorithm
4 The Outliers Selection Process 390
5 Main Results 391
6 Conclusion and Future Work 394
References 394
Robust Fuzzy Classification 395
1 Introduction 395
2 Robust Fuzzy Cluster Analysis 396
3 Confirmatory Analysis 398
References 402
Weighted Likelihood Inference for a Mixed Regressive Spatial Autoregressive Model 403
1 Introduction 403
2 A Weighted Likelihood Approach 404
3 A Small Simulation Study 406
4 A Real Example 408
5 Final Remarks 409
References 410
Detecting Price Outliers in European Trade Data with the Forward Search 411
1 Introduction 411
2 Application Context, Data and Statistical Patterns 412
3 Application of the Forward Search 413
4 Heuristic Comparison with the ``Backward'' Outliers 416
5 Towards an Automatic Procedure 416
6 Discussion and Main Conclusions 418
References 418
Part X Statistical Methods for Financial and Economics Data 420
Comparing Continuous Treatment Matching Methods in PolicyEvaluation 421
1 Introduction 421
2 Simulating a Subsidies Allocation Mechanism:The Case of L.488 422
3 The Matching Methods in the Continuous Framework 423
4 The Monte Carlo Experiment 424
5 Results and Conclusions 426
References 428
Temporal Aggregation and Closure of VARMA Models: Some New Results 429
1 Introduction 429
2 Temporal Aggregation 430
3 Temporal Aggregation and VARMA Models 431
4 ``Markovian" Representation 433
References 436
An Index for Ranking Financial Portfolios According to Internal Turnover 438
1 Introduction 438
2 Style Analysis Models 439
3 Ranking Portfolios According to Internal Turnover 441
4 Concluding Remarks 444
References 445
Bayesian Hidden Markov Modelsfor Financial Data 446
1 Introduction 446
2 The Model 447
3 Computational Implementation 448
4 Bayesian Inference and Forecasting 450
5 An Application to Financial Data 451
6 Conclusions 453
References 453
Part XI Missing Values 455
Regression Imputation for Space-Time Datasets with MissingValues 456
1 Introduction 456
2 A New Regression Single Imputation Method 458
3 Missing Data Simulation 459
4 Results 461
5 Toward a Multiple Imputation Method 461
References 463
A Multiple Imputation Approach in a Survey on University Teaching Evaluation 464
1 Introduction 464
2 An Imputation Procedure to Recover for Missingness 465
3 An Application to the Data on the Evaluationof University Teaching 466
3.1 AD 468
3.2 AE 468
4 Some Final Remarks 472
References 472