There has been astounding growth in the use of multivariable analysis in clinical research. When the first edition of this book was published in 1999 logistic regression and proportional hazards models were cutting-edge techniques. Now for many researchers, these are old, staid models and the new edge is mixed-effects models, generalized estimating equations, Poisson regression, and propensity score analysis.

The use of these more sophisticated models is fueled by the development of user-friendly software for constructing multivariable models, increased availability of electronic databases (medical records, disease and procedure registries) that provide longitudinal data on large populations, and increased funding for and interest in clinical effectiveness studies – studies comparing different treatments in use – as a method of improving quality and reducing healthcare costs.

What hasn't changed in the past 11 years is the need for an easy-to-follow guide for nonstatisticians on how to perform and interpret these models. Although the available software (e.g., SPSS, SAS, S-plus, R) doesn't require programming experience or mathematical aptitude to conduct the analyses, if the analysis is not set up correctly, the answer is sure to be wrong! Even when the analysis is performed correctly, researchers may not draw the correct conclusions from the output.

To prevent these problems, throughout the book I have focused on how to set up and interpret multivariable analysis. I use examples from the medical and public health literature because illustrations of how to correctly analyze data and present the results will help you analyze and present your data correctly.