(Ebook) Statistical learning for biomedical data by James D Malley; Karen G Malley; Sinisa Pajevic ISBN 9780521699099, 9780521875806, 0521699096, 0521875803

(Ebook) Statistical learning for biomedical data by James D Malley; Karen G Malley; Sinisa Pajevic ISBN 9780521699099, 9780521875806, 0521699096, 0521875803

This book is for anyone who has biomedical data and needs to identify variables that predict an outcome, for two-group outcomes such as tumor/not-tumor, survival/death, or response from treatment. Statistical learning machines are ideally suited to these types of prediction problems, especially if the variables being studied may not meet the assumptions of traditional techniques. Learning machines come from the world of probability and computer science but are not yet widely used in biomedical research. This introduction brings learning machine techniques to the biomedical world in an accessible way, explaining the underlying principles in nontechnical language and using extensive examples and figures. The authors connect these new methods to familiar techniques by showing how to use the learning machine models to generate smaller, more easily interpretable traditional models. Coverage includes single decision trees, multiple-tree techniques such as Random Forests(TM), neural nets, support vector machines, nearest neighbors and boosting Machine generated contents note: 1.Prologue -- 1.1.Machines that learn -- some recent history -- 1.2.Twenty canonical questions -- 1.3.Outline of the book -- 1.4.A comment about example datasets -- 1.5.Software -- Note -- 2.The landscape of learning machines -- 2.1.Introduction -- 2.2.Types of data for learning machines -- 2.3.Will that be supervised or unsupervised? -- 2.4.An unsupervised example -- 2.5.More lack of supervision -- where are the parents? -- 2.6.Engines, complex and primitive -- 2.7.Model richness means what, exactly? -- 2.8.Membership or probability of membership? -- 2.9.A taxonomy of machines? -- 2.10.A note of caution -- one of many -- 2.11.Highlights from the theory -- Notes -- 3.A mangle of machines -- 3.1.Introduction -- 3.2.Linear regression -- 3.3.Logistic regression -- 3.4.Linear discriminant -- 3.5.Bayes classifiers [-] regular and naive -- 3.6.Logic regression -- 3.7.k-Nearest neighbors -- 3.8.Support vector machines -- 3.9.Neural networks -- 3.10.Boosting -- 3.11.Evolutionary and genetic algorithms -- Notes -- 4.Three examples and several machines -- 4.1.Introduction -- 4.2.Simulated cholesterol data -- 4.3.Lupus data -- 4.4.Stroke data -- 4.5.Biomedical means unbalanced -- 4.6.Measures of machine performance -- 4.7.Linear analysis of cholesterol data -- 4.8.Nonlinear analysis of cholesterol data -- 4.9.Analysis of the lupus data -- 4.10.Analysis of the stroke data -- 4.11.Further analysis of the lupus and stroke data -- Notes -- 5.Logistic regression -- 5.1.Introduction -- 5.2.Inside and around the model -- 5.3.Interpreting the coefficients -- 5.4.Using logistic regression as a decision rule -- 5.5.Logistic regression applied to the cholesterol data -- 5.6.A cautionary note -- 5.7.Another cautionary note -- 5.8.Probability estimates and decision rules -- 5.9.Evaluating the goodness-of-fit of a logistic regression model -- 5.10.Calibrating a logistic regression -- 5.11.Beyond calibration -- 5.12.Logistic regression and reference models -- Notes -- 6.A single decision tree -- 6.1.Introduction -- 6.2.Dropping down trees -- 6.3.Growing a tree -- 6.4.Selecting features, making splits -- 6.5.Good split, bad split -- 6.6.Finding good features for making splits -- 6.7.Misreading trees -- 6.8.Stopping and pruning rules -- 6.9.Using functions of the features -- 6.10.Unstable trees? -- 6.11.Variable importance -- growing on trees? -- 6.12.Permuting for importance -- 6.13.The continuing mystery of trees -- 7.Random Forests -- trees everywhere -- 7.1.Random Forests in less than five minutes -- 7.2.Random treks through the data -- 7.3.Random treks through the features -- 7.4.Walking through the forest -- 7.5.Weighted and unweighted voting -- 7.6.Finding subsets in the data using proximities -- 7.7.Applying Random Forests to the Stroke data -- 7.8.Random Forests in the universe of machines -- Notes -- 8.Merely two variables -- 8.1.Introduction -- 8.2.Understanding correlations -- 8.3.Hazards of correlations -- 8.4.Correlations big and small -- Notes -- 9.More than two variables -- 9.1.Introduction -- 9.2.Tiny problems, large consequences -- 9.3.Mathematics to the rescue? -- 9.4.Good models need not be unique -- 9.5.Contexts and coefficients -- 9.6.Interpreting and testing coefficients in models -- 9.7.Merging models, pooling lists, ranking features -- Notes -- 10.Resampling methods -- 10.1.Introduction -- 10.2.The bootstrap -- 10.3.When the bootstrap works -- 10.4.When the bootstrap doesn't work -- 10.5.Resampling from a single group in different ways -- 10.6.Resampling from groups with unequal sizes -- 10.7.Resampling from small datasets -- 10.8.Permutation methods -- 10.9.Still more on permutation methods -- Note -- 11.Error analysis and model validation -- 11.1.Introduction -- 11.2.Errors? What errors? -- 11.3.Unbalanced data, unbalanced errors -- 11.4.Error analysis for a single machine -- 11.5.Cross-validation error estimation -- 11.6.Cross-validation or cross-training? -- 11.7.The leave-one-out method -- 11.8.The out-of-bag method -- 11.9.Intervals for error estimates for a single machine -- 11.10.Tossing random coins into the abyss -- 11.11.Error estimates for unbalanced data -- 11.12.Confidence intervals for comparing error values -- 11.13.Other measures of machine accuracy -- 11.14.Benchmarking and winning the lottery -- 11.15.Error analysis for predicting continuous outcomes -- Notes -- 12.Ensemble methods [--] let's take a vote -- 12.1.Pools of machines -- 12.2.Weak correlation with outcome can be good enough -- 12.3.Model averaging -- Notes -- 13.Summary and conclusions -- 13.1.Where have we been? -- 13.2.So many machines -- 13.3.Binary decision or probability estimate? -- 13.4.Survival machines? Risk machines? -- 13.5.And where are we going?