Librería Portfolio

A comprehensive introduction to statistical methods for data mining and knowledge discovery.

Applications of data mining and 'big data' increasingly take center stage in our modern, knowledge-driven society, supported by advances in computing power, automated data acquisition, social media development and interactive, linkable internet software. This book presents a coherent, technical introduction to modern statistical learning and analytics, starting from the core foundations of statistics and probability. It includes an overview of probability and statistical distributions, basics of data manipulation and visualization, and the central components of standard statistical inferences. The majority of the text extends beyond these introductory topics, however, to supervised learning in linear regression, generalized linear models, and classification analytics. Finally, unsupervised learning via dimension reduction, cluster analysis, and market basket analysis are introduced.

Extensive examples using actual data (with sample R programming code) are provided, illustrating diverse informatic sources in genomics, biomedicine, ecological remote sensing, astronomy, socioeconomics, marketing, advertising and finance, among many others.

Statistical Data Analytics:


Focuses on methods critically used in data mining and statistical informatics. Coherently describes the methods at an introductory level, with extensions to selected intermediate and advanced techniques.
Provides informative, technical details for the highlighted methods.
Employs the open-source R language as the computational vehicle - along with its burgeoning collection of online packages - to illustrate many of the analyses contained in the book.
Concludes each chapter with a range of interesting and challenging homework exercises using actual data from a variety of informatic application areas.

This book will appeal as a classroom or training text to intermediate and advanced undergraduates, and to beginning graduate students, with sufficient background in calculus and matrix algebra. It will also serve as a source-book on the foundations of statistical informatics and data analytics to practitioners who regularly apply statistical learning to their modern data.



Table of Contents

Preface xiii
Part I Background: Introductory Statistical Analytics 1

1 Data analytics and data mining 3

1.1 Knowledge discovery: finding structure in data 3

1.2 Data quality versus data quantity 5

1.3 Statistical modeling versus statistical description 7

2 Basic probability and statistical distributions 10

2.1 Concepts in probability 10

2.1.1 Probability rules 11

2.1.2 Random variables and probability functions 12

2.1.3 Means, variances, and expected values 17

2.1.4 Median, quartiles, and quantiles 18

2.1.5 Bivariate expected values, covariance, and correlation 20

2.2 Multiple random variables* 21

2.3 Univariate families of distributions 23

2.3.1 Binomial distribution 23

2.3.2 Poisson distribution 26

2.3.3 Geometric distribution 27

2.3.4 Negative binomial distribution 27

2.3.5 Discrete uniform distribution 28

2.3.6 Continuous uniform distribution 29

2.3.7 Exponential distribution 29

2.3.8 Gamma and chi-square distributions 30

2.3.9 Normal (Gaussian) distribution 32

2.3.10 Distributions derived from normal 37

2.3.11 The exponential family 41

3 Data manipulation 49

3.1 Random sampling 49

3.2 Data types 51

3.3 Data summarization 52

3.3.1 Means, medians, and central tendency 52

3.3.2 Summarizing variation 56

3.3.3 Summarizing (bivariate) correlation 59

3.4 Data diagnostics and data transformation 60

3.4.1 Outlier analysis 60

3.4.2 Entropy* 62

3.4.3 Data transformation 64

3.5 Simple smoothing techniques 65

3.5.1 Binning 66

3.5.2 Moving averages* 67

3.5.3 Exponential smoothing* 69

4 Data visualization and statistical graphics 76

4.1 Univariate visualization 77

4.1.1 Strip charts and dot plots 77

4.1.2 Boxplots 79

4.1.3 Stem-and-leaf plots 81

4.1.4 Histograms and density estimators 83

4.1.5 Quantile plots 87

4.2 Bivariate and multivariate visualization 89

4.2.1 Pie charts and bar charts 90

4.2.2 Multiple boxplots and QQ plots 95

4.2.3 Scatterplots and bubble plots 98

4.2.4 Heatmaps 102

4.2.5 Time series plots* 105

5 Statistical inference 115

5.1 Parameters and likelihood 115

5.2 Point estimation 117

5.2.1 Bias 118

5.2.2 The method of moments 118

5.2.3 Least squares/weighted least squares 119

5.2.4 Maximum likelihood* 120

5.3 Interval estimation 123

5.3.1 Confidence intervals 123

5.3.2 Single-sample intervals for normal (Gaussian) parameters 124

5.3.3 Two-sample intervals for normal (Gaussian) parameters 128

5.3.4 Wald intervals and likelihood intervals* 131

5.3.5 Delta method intervals* 135

5.3.6 Bootstrap intervals* 137

5.4 Testing hypotheses 138

5.4.1 Single-sample tests for normal (Gaussian) parameters 140

5.4.2 Two-sample tests for normal (Gaussian) parameters 142

5.4.3 Walds tests, likelihood ratio tests, and 'exact' tests* 145

5.5 Multiple inferences* 148

5.5.1 Bonferroni multiplicity adjustment 149

5.5.2 False discovery rate 151

Part II Statistical Learning and Data Analytics 161

6 Techniques for supervised learning: simple linear regression 163

6.1 What is "supervised learning?ö 163

6.2 Simple linear regression 164

6.2.1 The simple linear model 164

6.2.2 Multiple inferences and simultaneous confidence bands 171

6.3 Regression diagnostics 175

6.4 Weighted least squares (WLS) regression 184

6.5 Correlation analysis 187

6.5.1 The correlation coefficient 187

6.5.2 Rank correlation 190

7 Techniques for supervised learning: multiple linear regression 198

7.1 Multiple linear regression 198

7.1.1 Matrix formulation 199

7.1.2 Weighted least squares for the MLR model 200

7.1.3 Inferences under the MLR model 201

7.1.4 Multicollinearity 208

7.2 Polynomial regression 210

7.3 Feature selection 211

7.3.1 R2p plots 212

7.3.2 Information criteria: AIC and BIC 215

7.3.3 Automated variable selection 216

7.4 Alternative regression methods* 223

7.4.1 Loess 224

7.4.2 Regularization: ridge regression 230

7.4.3 Regularization and variable selection: the Lasso 238

7.5 Qualitative predictors: ANOVA models 242

8 Supervised learning: generalized linear models 258

8.1 Extending the linear regression model 258

8.1.1 Nonnormal data and the exponential family 258

8.1.2 Link functions 259

8.2 Technical details for GLiMs* 259

8.2.1 Estimation 260

8.2.2 The deviance function 261

8.2.3 Residuals 262

8.2.4 Inference and model assessment 264

8.3 Selected forms of GLiMs 265

8.3.1 Logistic regression and binary-data GLiMs 265

8.3.2 Trend testing with proportion data 271

8.3.3 Contingency tables and log-linear