Intrenion

How Not to Be Wrong (Jordan Ellenberg)

Table of Contents

Copy Doctrine

Practice 1: Normalize numbers before comparing groups

Problem
Raw totals can create misleading comparisons.

Action
Adjust measurements to a common scale before comparing them.

Outcome
You make more accurate judgments.

Chapter: Linearity - Less Like Sweden

Practice 2: Limit linear estimates to the range where they fit

Problem
A trend that looks straight nearby may bend over larger ranges.

Action
Check whether the relationship changes shape before extending it.

Outcome
You avoid faulty predictions.

Chapter: Linearity - Straight Locally, Curved Globally

Practice 3: Examine the distribution behind the average

Problem
An average can hide important differences within a group.

Action
Look at how values are spread across the population.

Outcome
You understand the data more accurately.

Chapter: Linearity - Everyone Is Obese

Practice 4: Convert numbers into familiar units

Problem
Large numbers are difficult to understand on their own.

Action
Express quantities using a reference people can easily imagine.

Outcome
You judge scale more clearly.

Chapter: Linearity - How Much Is That in Dead Americans?

Practice 5: Compare quantities to the whole they belong to

Problem
A quantity can appear important without context.

Action
Measure each part as a share of the total.

Outcome
You assess significance more accurately.

Chapter: Linearity - More Pie Than Plate

Practice 6: Consider chance before assuming a pattern is meaningful

Problem
Random events can create convincing patterns.

Action
Ask how likely the pattern is to appear by chance.

Outcome
You avoid false discoveries.

Chapter: Inference - The Baltimore Stockbroker and the Bible Code

Practice 7: Verify surprising findings with proper tests

Problem
Random noise can look like a real effect.

Action
Check whether the result survives controls and repeated testing.

Outcome
You identify genuine evidence more reliably.

Chapter: Inference - Dead Fish Don’t Read Minds

Practice 8: Demand stronger evidence for unlikely claims

Problem
Unusual conclusions can be supported by coincidence.

Action
Require evidence that clearly exceeds what randomness would produce.

Outcome
You reach more reliable conclusions.

Chapter: Inference - Reductio Ad Unlikely

Practice 9: Evaluate the method before trusting the result

Problem
Weak methods can produce convincing but false findings.

Action
Examine how the evidence was collected and tested.

Outcome
You distinguish reliable results from unreliable ones.

Chapter: Inference - The International Journal of Haruspicy

Practice 10: Update beliefs when new evidence arrives

Problem
Ignoring prior knowledge or new information leads to poor estimates.

Action
Combine existing knowledge with the latest evidence.

Outcome
You make more accurate judgments under uncertainty.

Chapter: Inference - Are You There, God? It’s Me, Bayesian Inference

Practice 11: Judge choices by their expected value

Problem
Rare rewards can distract attention from poor odds.

Action
Calculate the average outcome across all possible results.

Outcome
You make better decisions under uncertainty.

Chapter: Expectation - What to Expect When You’re Expecting to Win the Lottery

Practice 12: Optimize for long-term results rather than perfect outcomes

Problem
Trying to avoid every mistake can reduce overall performance.

Action
Choose the option with the best average result over time.

Outcome
You achieve better long-term outcomes.

Chapter: Expectation - Miss More Planes!

Practice 13: Find where competing effects balance each other

Problem
Opposing forces can make outcomes difficult to understand.

Action
Identify the point where the effects become equal.

Outcome
You understand the system more clearly.

Chapter: Expectation - Where the Train Tracks Meet

Practice 14: Expect extreme results to move toward the average

Problem
Exceptional performance often contains a large element of chance.

Action
Adjust future expectations to a more typical level.

Outcome
You make more realistic forecasts.

Chapter: Regression - The Triumph of Mediocrity

Practice 15: Focus on the underlying trend instead of outliers

Problem
Individual observations can be distorted by random variation.

Action
Use the central pattern in the data to guide estimates.

Outcome
You make more dependable predictions.

Chapter: Regression - Galton’s Ellipse

Practice 16: Separate causation from correlation

Problem
Variables can move together without causing each other.

Action
Look for evidence that one factor directly changes the other.

Outcome
You avoid mistaken explanations.

Chapter: Regression - Does Lung Cancer Make You Smoke Cigarettes?

Practice 17: Define what exists before measuring it

Problem
Vague concepts cannot be analyzed consistently.

Action
Specify clearly what is being studied and how it is identified.

Outcome
You build stronger conclusions.

Chapter: Regression - Part V: Existence

Practice 18: Treat public opinion as a range of views

Problem
A single average can hide important disagreement.

Action
Examine how opinions vary across people and groups.

Outcome
You gain a more accurate picture of collective beliefs.

Chapter: Regression - There Is No Such Thing as Public Opinion

Practice 19: Study the rules that generate complex patterns

Problem
Complex outcomes can seem mysterious when only the results are observed.

Action
Investigate the simple processes that create the observed behavior.

Outcome
You understand complex systems more deeply.

Chapter: Regression - “Out of Nothing I Have Created a Strange New Universe”