Relevant For:
- Researchers who run experiments or analyze observational data
- Graduate students and early-career scientists learning statistical judgment
- Data practitioners who want robust decision rules, not rituals
Remarkable Points:
- Converts abstract inferential errors into short, actionable checks
- Places experimental design and question formulation before statistical toys
- Shows how selective reporting and multiple comparisons create false narratives
- Makes reproducibility a routine, not an optional badge
Most Liked (As-Is Statements):
"Ask the question you will be able to answer."
"P values do not measure the probability that your hypothesis is true."
Key Themes:
Theme 1: Design Before Analysis
Reinhart insists that a redundant or vague question dooms inference. The clearest gains come from predefining comparisons and stopping rules.
Theme 2: Treat P Values as Diagnostics
P values are signals, not verdicts. The book reframes them as one input among many and warns against binary thinking.
Theme 3: Guardrails for Reporting
Transparent code, full data, and honest reporting are presented as procedural safeguards that change incentives and reduce error.
Theme 4: Simple Models, Honest Claims
Prefer models that match the question over complex fits that seduce with precision. State assumptions clearly and avoid overreach.
Final Remark:
Statistics Done Wrong is a corrective manual: short, direct, and tactical. It replaces ritual with rules that scale across fields. The value is not new mathematics but better habits. Read it to shrink the gap between what analyses show and what claims deserve.
Clarity in statistical practice starts with asking the right question and sticking to the plan.
Similar Books:
- The Art of Statistics by David Spiegelhalter
- Naked Statistics by Charles Wheelan
- The Signal and the Noise by Nate Silver
- Practical Statistics for Data Scientists by Peter Bruce and Andrew Bruce