Regression Methods in Biostatistics

2008-02-10

This book is about regression techniques commonly used when modelling continuous and/or binary outcomes in biomedical studies. Starting from the most basic techniques (but too often neglected, to my opinion) of exploratory and descriptive techniques (Chap. 2, graphical and numerical summaries), the authors devote an entire chapter (Chap. 3) to give the reader a clear overview of classical multivariate techniques used to characterize association between categorical and continuous variable (includind censored data). Next, they provide an in-depth coverage of each of these methods in separate chapter.

This includes:

  • Chapter 4: Linear regression. Reviewing the basic properties of the multiple linear regression model; Testing for trend and association between multiple categorical predictors; Discussion about confounding, mediation and interaction (rarely found in other textbooks!); Checking model assumption (linearity, normality, homoscedasticity, outlying, transformation) and model fit./li>
  • Chapter 5: Predictor selection. Causal model and rules (e.g. RMSE, CART) for selecting single or multiple predictor of interest; Dealing with collinearity and Backward selection; Assessing model selection./li>
  • Chapter 6: Logisitic regression. Use of single and multiple predictors with a binary response; Case-control studies and matched pairs analysis; Checking logistic model fit; Alternatives to standard logistic regression (including other response variable like relative risk, nonparametric model, ordinal outcome)./li>
  • Chapter 7: Survival analysis. Overview of survival data; Cox proportional hazards model (parametric vs. semi-parametric model, binary vs. multiple response outcome, categorical vs. continuous predictor) and its extension (time-depedent covariates and stratification); Checking model fit; Other details related to censored data (bootstraped confidence interval, independent or interval censoring, left-censoring)./li>
  • Chapter 8: Repeated measures analysis. Hierarchical and longitudinal design; GEE approach; Random-effects models (including prediction with binary response); Deriving confidence intervals by bootstrap./li>
  • Chapter 9: Generalized linear models. Three case studies illustrating the general approach when using an GLM (specifying a linear combination of predictors related to the mean response, choosing an appropriate link function, accounting for variance-mean relationship, interpreting the parameters estimates and model fit)./li>
  • Chapter 10: Complex surveys. A nice introduction to sampling issues and survey methodology; Probability weighting and variance estimation in classical survey design.

This really is a must-have one book for every practician interested in regression modelling for the life sciences.

---

Articles with the same tag(s):

Academic teaching
Data cleaning techniques
Data Science from Scratch
Writing a book
Stata for health researchers
R Graphs Cookbook
Bad Data
Data science at the command-line
Reproducible research with R
Twenty canonical questions in machine learning

---