Where Does 1/√n Shrinkage Show Up?
The standard error of a mean shrinks as 1/√n, and this same scaling shows up visibly in dice averages, polling margins, and physics timing experiments.
A complete interactive classroom, not just a preview.
Start when you are ready to enter this Stage's 9 scenes and explore, respond, and learn as you go.
Where does the 1/√n shrinkage of standard error actually appear in real measurements?
Polls, dice, and coin flips all get more accurate as you collect more data — but how fast?
Intuition says doubling the data halves the noise, yet the math predicts it only shrinks by roughly 30%.
Side-by-side comparison of how standard error falls with √n across dice rolls, poll samples, and physics timing experiments.
The 1/√n law is not a textbook abstraction — it appears in everyday measurements whenever we average many independent observations.
If I double the number of samples, the error probably halves too.
- Derivation of the central limit theorem
- Confidence interval formulas beyond a qualitative margin
- Bias, non-Gaussian errors, and correlated samples
- 01The Slow Shrinkage PuzzleslideQuestion
Frame the driving question: when we average more measurements, how quickly does our uncertainty shrink?
- Averaging N independent measurements reduces noise
- Intuition suggests error shrinks in proportion to N
- The real rule is far weaker than that
- 02Your Intuition CheckquizPrediction
Commit to how you think standard error scales with the number of samples.
- Make one independent choice about the scaling law
- 03Dice Average SimulatorinteractiveEvidence
Roll 1 to 400 dice and watch how the running average of a six-sided die stabilizes around 3.5.
- Roll a chosen number of dice and read the running mean
- Compare spread at n=10 vs n=100 vs n=400
- Notice that spread shrinks slowly, not linearly
- 04Poll Margin ComparisoninteractiveEvidence
Compare reported margins of error for hypothetical polls of 100, 400, 1600, and 6400 respondents.
- Adjust the sample size with a slider
- Read the implied margin of error for each size
- See the √n scaling directly in the labels
- 05Why 1/√n, Not 1/nslideExplanation
Explain why averaging N independent measurements with standard deviation σ gives a standard error of σ/√N.
- Variances add, so the variance of the mean is σ²/N
- Standard error is the square root: σ/√N
- Quadrupling the data only halves the error
- 06Physics Stopwatch TestinteractiveEvidence
Time how long it takes a pendulum to complete 10 swings, repeat the measurement, and watch how the standard error shrinks as you add trials.
- Adjust the number of repeated timing trials
- Read the standard error reported for each run count
- Confirm the √n curve against your own hand-collected data
- 07When the Law BreaksslideBoundary
Show where 1/√n shrinkage no longer applies: biased instruments, correlated samples, and heavy-tailed errors.
- Systematic bias is not reduced by averaging
- Autocorrelated measurements have effective N, not raw N
- Outliers from heavy-tailed distributions slow convergence
- 08Apply It: Lab Report PlannerinteractiveTransfer
Given a desired precision for a reaction-time measurement, decide how many trials to run.
- Enter your target standard error in milliseconds
- Read the suggested trial count
- See how tripling precision requires nine times the data
- 09Answering the Driving QuestionslideResolution
Resolve the opening tension by naming the three real measurements where 1/√n shrinkage is visibly at work.
- Dice averages: visible narrowing around 3.5
- Polls: 1/√n sets the printed margin of error
- Physics timing: standard error of repeated trials follows the same curve
Discussion threads for a Stage aren't available yet.
