Relevant For:
- Advanced undergraduates and graduate students needing a principled pathway from t tests to GLMs
- Researchers in ecology, evolution, and environmental science who want reproducible, interpretation-focused workflows
- Practitioners replacing p-value rituals with estimation and model comparison
Remarkable Points:
- Prioritizes estimation and interval thinking over binary testing
- Uses information criteria as a coherent currency for model comparison
- Organizes diagnostics and model building into a compact, repeatable sequence
- Road-tested examples in R that slot into a reproducible analysis pipeline
Most Liked (As-Is Statements):
"Treat parameters as quantities to estimate, not booleans to decide."
"Model comparison is a relative currency, not an absolute verdict."
Key Themes:
Theme 1: Estimation First
The book replaces hypothesis ritual with interval and effect-size reasoning. That shift turns uncertainty into a description to be interpreted, not a gatekeeper to be feared.
Theme 2: Models as Tools, Not Answers
Models are presented as provisional maps. The emphasis is on comparing alternative maps and understanding where each one breaks down.
Theme 3: Information Criteria as Decision Framework
AIC and related measures are treated as practical decision devices for model choice, integrating fit and parsimony into a single, interpretable metric.
Theme 4: Workflow and Reproducibility in R
R code is not incidental; it encodes the analytic discipline. The worked examples teach a compact, repeatable pipeline from data to interpretation.
Final Remark:
The New Statistics with R succeeds not by covering every method but by clarifying how to think about models, estimates, and uncertainty. It provides mental models that collapse common analysis mistakes into predictable steps. Readers come away with a procedural habit: prefer estimates, compare models, diagnose failures, then communicate effect sizes. For its intended disciplines the book is a practical instrument to sharpen statistical judgment rather than a catalog of techniques.
Clarity of thought is the book's primary deliverable; the R code is the mechanism that makes that clarity repeatable.
Similar Books:
- Statistical Rethinking by Richard McElreath
- Model Selection and Multimodel Inference by Burnham and Anderson
- Data Analysis Using Regression and Multilevel/Hierarchical Models by Gelman and Hill