It’s a strong statement to declare something the “bible” of a particular subject, but this text comes close with statistics. It touches on almost all major topics in general linear modeling, and covers some topics in experimental design as well. This book was used in my graduate-level linear regression course, but this text would be perfectly suitable for a two-semester undergraduate course as well. Aimed at practitioners, statisticians, data scientists, and engineers, it’s highly readable, with provided datasets for the exercises, and a rigorous treatment of the topics that doesn’t wander too far into the abstract. The text even is supplemented with both an instructor and a student solutions manual, which makes this an excellent self-study book. It takes the time to discuss linear model diagnostics, remedial actions, and building the model for simple linear regression, multiple linear regression, and regression involving both quantitative and qualitative predictors. In particular, I like that it discusses extra sums of squares tests for whether or not several terms in the regression model can be dropped, discussing the issue with stepwise model selection (though not in the detail I’d prefer). It touches on experimental design and neural networks, but from the analysis perspective rather than the design perspective. Occasionally, derivations are done, such as the regression coefficients using the normal equations, but there aren’t really many proofs. The text is meant to develop a statistical sense, and the ability to deal with data. I consider it a necessary read for anyone looking to become or currently working as a data scientist, as the nuances of seemingly simple general linear regression can become lost with the ease of packaged software and implementation.

-Rachel Traylor, Ph.D.

### Review

## Prerequisites

### solid grasp of and comfort with conventional algebra, some basic statistical concepts, calculus is a plus but not necessary

## Topics Covered

### Part 1: Simple Linear Regression

- linear regression with one predictor
- inference in regression and correlation analysis
- diagnostics and remedial measures
- simultaneous inferences

### Part 2: Multiple Linear Regression

- general linear model in matrix terms
- ANOVA results
- extra sums of squares and their use in regression tests
- dealing with qualitative and quantitative predictors
- model selection and validation, diagnostics, and remedial measures
- intro to time series data: autocorrelation

### Part 3: Nonlinear Regression

- intro to nonlinear regression and neural networks
- logistic and poisson regression

### Part 4: Design and Analysis of Single Factor Studies

- single-factor studies
- analysis of factor level means
- ANOVA diagnostics

### Part 5: Multi-Factor Studies

- Two-Factor studies
- randomized complete block designs
- ANCOVA
- Multi-factor studies
- Random and Mixed Effects Models

### Part 6: Specialized Study Designs

- nested designs
- released measure designs
- balanced incomplete block designs
- Factorial and fractional factorial designs
- response surface methodology

### Attributes

Difficulty | 2 |

Good for teaching? | 5 |

Proof Quality | NA |

Quality of Exercises | 5 |

Readability | 5 |

Self-Study | 4 |