Complex Data Modeling and Computationally Intensive Statistical Methods (eBook)

Pietro Mantovan, Piercesare Secchi (Autoren)

eBook Download: PDF

2011 | 2010
X, 164 Seiten
Springer Italia (Verlag)
978-88-470-1386-5 (ISBN)

Infos

Selected from the conference 'S.Co.2009: Complex Data Modeling and Computationally Intensive Methods for Estimation and Prediction,' these 20 papers cover the latest in statistical methods and computational techniques for complex and high dimensional datasets.

Pietro Mantovan has been Professor of Statistics since 1986 at the University Ca' Foscari of Venezia, Italy, where he has served as coordinator of research units, head of the Departement of Statistics, and Dean of the Faculty of Economics. He has written several articles, monographs and textbooks on classical and Bayesian methods for statistical inference. His recent research interests focus on Bayesian methods for learning and prediction, statistical perturbation models for matrix data, dynamic regression with covariate errors, parallel algorithms for system identification in dynamic models, on line monitoring and forecasting of environmental data, hydrological forecasting uncertainty assessment, and robust inference processes.

Piercesare Secchi is Professor of Statistics at MOX since 2005 and Director of the Department of Mathematics at the Politecnico di Milano. He got a Doctorate in Methodological Statistics from the University of Trento in 1992 and a PhD in Statistics from the University of Minnesota in 1995. He has written several papers on stochastic games and on Bayesian nonparametric predictive inference and bootstrap techniques. His present research interests focus on statistical methods for the exploration, classification and analysis of high dimensional data, like functional data or images generated by medical diagnostic devices or by remote sensing. He also works on models for Bayesian inference, in particular those generated by urn schemes, on response adaptive designs of experiments for clinical trials and on biodata mining. He is PI of different projects in applied statistics and coordinator of the Statistical Unit of the Aneurisk project.

Inhalt

Title Page 1
Copyright Page 4
Preface 5
Table of Contents 7
List of Contributors 9
Space-time texture analysis in thermal infraredimaging for classification of Raynaud’s Phenomenon 11
1 Introduction 11
2 TheData 12
3 Processing thermal high resolution infrared images 13
3.1 Segmentation 13
3.2 Registration 13
4 Feature extraction 15
4.1 ST-GMRFs 16
4.2 Texture statistics through co-occurrence matrices 18
5 Classification results 19
6 Conclusions 20
References 21
Mixed-effects modelling of Kevlar fibre failure timesthrough Bayesian non-parametrics 23
1 Introduction 23
2 Accelerated life models for Kevlar fibre life data 25
3 The Bayesian semiparametric AFT model 26
4 Data analysis 28
5 Conclusions 34
Appendix 34
References 36
Space filling and locally optimal designs for Gaussian Universal Kriging 37
1 Introduction 37
2 Kriging methodology 39
3 Optimality of space filling designs 40
4 Locally optimal designs for Universal Kriging 41
4.1 Optimal designs for estimation 41
4.2 Optimal designs for prediction 46
5 Conclusions 48
References 48
Exploitation, integration and statistical analysis of thePublic Health Database and STEMI Archive in theLombardia region 50
1 Introduction 50
2 The MOMI2 study 52
3 The STEMI Archive 55
4 The Public Health Database 56
4.1 Healthcare databases 57
4.2 Health information systems in Lombardia 58
5 The statistical perspective 58
5.1 Frailty models 59
5.2 Generalised linear mixed models 60
5.3 Bayesian hierarchical models 61
6 Conclusions 62
References 62
Bootstrap algorithms for variance estimation in PS sampling 65
1 Introduction 65
2 The naïve boostrap 66
3 Holmberg’s PS bootstrap 67
4 The 0.5 PS-bootstrap 70
5 The x-balanced PS-bootstrap 70
6 Simulation study 71
7 Conclusions 76
References 76
Fast Bayesian functional data analysis of basal body temperature 78
1 Introduction 78
2 Methods 80
2.1 RVM in linear models 80
2.2 Extension to linear mixed model 81
3 Results: application to bbt data 84
3.1 Subject-specific profiles 85
3.2 Subject-specific and population average profiles 86
3.3 Prediction 88
4 Conclusions 88
References 89
A parametric Markov chain to model age- and state-dependent wear processes 91
1 Introduction 91
2 System description and preliminary technological considerations 93
3 Data description and preliminary statistical considerations 94
4 Model description 97
5 Parameter estimation 99
6 Testing dependence on time and/or state 101
7 Conclusions 102
References 103
Case studies in Bayesian computation using INLA 104
1 Introduction 104
2 Latent Gaussian models 105
3 Integrated Nested Laplace Approximation 107
4 The INLA package for R 108
5 Case studies 108
5.1 A GLMM with over-dispersion 108
5.2 Childhood under nutrition in Zambia: spatial analysis 110
5.3 A simple example of survival data analysis 115
6 Conclusions 117
References 118
A graphical models approach for comparing gene sets 120
1 Introduction 104
2 Latent Gaussian models 105
3 Integrated Nested Laplace Approximation 107
4 The INLA package for R 108
5 Case studies 108
5.1 A GLMM with over-dispersion 108
5.2 Childhood undernutrition in Zambia: spatial analysis 110
5.3 A simple example of survival data analysis 115
6 Conclusions 117
References 118
A graphical models approach for comparing gene sets 120
1 Introduction 120
2 A brief introduction to pathways 121
3 Data and graphical models setup 123
4 Test of equality of two concentration matrices 125
5 Conclusions 126
References 126
Predictive densities and prediction limits based onpredictive likelihoods 128
1 Introduction 128
2 Review on predictive methods 129
2.1 Plug-in predictive procedures and improvements 130
2.2 Profile predictive likelihood and modifications 131
3 Likelihood-based predictive distributions and prediction limits 132
3.1 Probability distributions from predictive likelihoods 133
3.2 Prediction limits and coverage probabilities 135
4 Examples 135
4.1 Prediction limits for the sum of future Gaussian observations 136
4.2 Prediction limits for the maximum of future Gaussian observations 138
Appendix 139
References 141
Computer-intensive conditional inference 142
1 Introduction 142
2 An inference problem 144
3 Exponential family and ancillary statistic models 145
4 Analytic approximations 146
5 Bootstrap approximations 147
6 Examples 149
6.1 Inverse Gaussian distribution 149
6.2 Log-normal mean 150
6.3 Weibull distribution 151
6.4 Exponential regression 152
7 Conclusions 153
References 154
Monte Carlo simulation methods for reliability estimation and failure prognostics 156
1 Introduction 157
2 The subset and line sampling methods for realiability estimation 158
3 Particle filtering for failure prognosis 161
4 Conclusions 166
References 167

Produkt-Details

• Erscheint lt. Verlag: 27.1.2011
• Reihe/Serie: Contributions to Statistics
• Reihe/Serie: Contributions to Statistics
• Zusatzinfo: X, 170 p.
• Verlagsort: Milano
• Sprache: englisch
• Themenwelt: Informatik ► Datenbanken ► Data Warehouse / Data Mining
• Themenwelt: Mathematik / Informatik ► Mathematik ► Computerprogramme / Computeralgebra
• Themenwelt: Mathematik / Informatik ► Mathematik ► Statistik
• Themenwelt: Mathematik / Informatik ► Mathematik ► Wahrscheinlichkeit / Kombinatorik
• Themenwelt: Technik
• Schlagworte: Bayesian Statistics • biodata mining • classification • Classification and prediction of high dimensional data • complex data surveys • computational methods for statistics • Data Analysis • Data Mining • likelihood • machine learning • Statistica • statistical methods for industry and technology • Statistics • Time Series • Variance
• ISBN-10: 88-470-1386-0 / 8847013860
• ISBN-13: 978-88-470-1386-5 / 9788847013865