Application of Regularized Regressions to Identify Novel Predictors in Clinical Research

This textbook is an important novel menu for multiple variables regression entitled "regularized regression". It is a must have for identifying unidentified leading factors. Also, you get fitted parameters for your overfitted data. Finally, there is no more need for commonly misunderstood p-values. Instead, the regression coefficient, R-value, as reported from a regression line has been applied as the key predictive estimator of the regression study. With simple one by one variable regression it is no wider than -1 to +1. With multiple variables regression it can easily get > +1 or < -1. This means we have a seriously flawed regression model, mostly due to collinearity or non-linear data. Completing the analysis will lead to overfitting, and thus a meaningless significant study due to data spread wider than compatible with random. In order for the regression coefficients to remain in the right size, fortunately a shrinking procedure has been invented. In the past two decades regularized regression has become a major topic of research, particularly with high dimensional data. Yet, the method is pretty new and infrequently used in real-data analysis. Its performance as compared to traditional null hypothesis testing has to be confirmed by prospective comparisons. Most studies published to date are of a theoretical nature involving statistical modeling and simulation studies. The journals Nature and Science published 19 and 10 papers of this sort in the past 8 years. The current edition will for the first time systematically test regularized regression against traditional regression analysis in 20 clinical data examples. The edition is also a textbook and tutorial for medical and healthcare students as well as recollection bench and help desk for professionals. Each chapter can be studied as a standalone, and, using, real as well as hypothesized data, it tests the performance of the novel methodology against traditional regressions. Step by step analyses of 20 data files are included for self-assessment. The authors are well qualified in their field. Professor Zwinderman is past-president of the International Society of Biostatistics and Professor Cleophas is past-president of the American College of Angiology. The authors have been working together for 25 years and their research can be characterized as a continued effort to demonstrate that clinical data analysis is a discipline at the interface of biology and mathematics.

Table of Contents:
.- Basic Principles of Regression Analysis. .- Optimal Scaling, Discretization, and Regularization vs Traditional Linear Regression. .- Regularized Regression Analysis, Ridge, Lasso, Elastic Net Regression Coefficients. .- Effect of Predictors on Health Scores, 110 Patients, Traditional vs Regularized Regressions. .- Effect on Physical strength of Races, 60 Patients, Traditional Regression vs Regularized regressions. .- Effects of Genetic Polymorphisms on Clinical Outcomes, 250 Patients, Traditional vs Regularized Regressions. .- Effect of Old Treatment and Age on New Treatment, 35 Patients, Traditional vs Regularized Regressions. .- Effect on Paroxysmal Atrial Fibrillations of Four Predictors, 50 Patients, Traditional vs Regularized Regressions. .- Effect of Air Quality of Operating Rooms on Infections, 8 Operating Rooms, Traditional vs Regularized Regressions. .- Effect on Weightloss of Age, Calorieintake, Exercise, Interaction, 64 Patients, Traditional vs Regularized Regressions. .- Effect on Body Surface Measured of Gender, Age, Weight, Height, and Weight x Height Interaction, 90 Patients, Traditional vs Regularized Regressions. .- Effect on Paroxysmal Atrial Fibrillations of Gender, Treatment and Their Interaction, 40 Patients, Traditional vs Regularized Regressions. .- Effect on Hours of Sleep of Treatment Group, Age, Gender, Comorbidity, 20 Patients, Traditional vs Regularized Regressions. .- Effect of Betaagonist and Prednisone on Peak Expiratory Flow, 78 COPD Patients, Traditional vs Regularized Regressions. .- Effect on LDL Cholesterol Reduction of Five Predictors, 953 Patients, Traditional vs Regularized Regressions. .- Effect of Five Factors on Body Weight, 217 Patients, Traditional vs Regularized Regressions. .- Functional Data Analysis and Regularized Regressions.

About the Author :

Ton J. Cleophas is internist-clinical pharmacologist at the Department of Medicine Albert Schweitzer Hospital Dordrecht the Netherlands. He is also professor of Statistics and member of the Scientific Committee of the European College of Pharmaceutical Medicine Lyon France. He is particularly interested in machine learning methodologies and published many complete-overview-textbooks of the subject.

Aeilko H. Zwinderman is professor of Statistics and Chair of the Department of Biostatistics and Epidemiology at the University of Amsterdam the Netherlands. His current work focuses on development and validation of multivariable models, particularly in genetic research, and he is a major developer of penalized canonical analysis.