A Guide to Robust Statistical Methods

Robust statistical methods are now being used in a wide range of disciplines. The appeal of these methods is that they are designed to perform about as well as classic techniques when standard assumptions are truebut they continue to perform well in situations where classic methods perform poorly.

This book provides a relatively non-technical guide to modern methods. The focus is on applying modern methods using R, understanding when and why classic methods can be unsatisfactory, and fostering a conceptual understanding of the relative merits of different techniques. A recurring theme is that no single method reveals everything one would like to know about the population under study. An appeal of robust methods is that under general conditions they provide much higher power than conventional techniques. Perhaps more importantly, they help provide a deeper and more nuanced understanding of data.

The book is for readers who had at least one semester of statistics, aimed at non-statisticians.

Examines properties of robust estimators and their relative merits Focuses on heteroscedastic techniques, including recent advances dealing multicollinearity Contains recent advances dealing with measures effect size and outliers, including bad leverage points

Autorentext

Rand R. Wilcox is Professor of Psychology, USC Dornsife College of Letters, Arts and Sciences. He has written 15 other statistics books, including Fundamentals of Modern Statistical Methods: Substantially Improving Power and Accuracy, 2nd edition (2010), over 400 journal articles, is a former associate editor for 5 statistics journals, is an elected member of the International Statistical Institute and he created the R package, WRS.

Inhalt

1. Introduction.- 2. The one-sample case.- 3. Comparing two independent groups.- 4. Comparing two dependent groups.- 5. Comparing multiple independent groups.- 6. Comparing multiple dependent groups.- 7. Robust regression estimators.- 8. Inferential methods based on robust regression estimators.- 9. Measures of association.- 10. Comparing groups when there is a covariate.