Ebrahim AlhamedFrameworks Library

m.01 · I · Quantifying Uncertainty · Sampling & Confidence Intervals

The Range You Can Defend

Central Limit Theorem · Confidence Intervals · Standard Error.

The core idea

Every data-based decision rests on a sample. Every sample has noise. The central question of data analysis is not "what is the answer?" but "how wide is the range of plausible answers?" Confidence intervals give you that range in a defensible form — the standard error shrinks with sample size, and the interval widens with the confidence you demand. — after the Central Limit Theorem

The hero diagram

The sampling distribution.

As n grows, the distribution of sample means tightens around the true population mean. That tightening is the standard error.

Bell-curve family Three overlapping bell curves sharing the same center (true mean) but progressively narrower, illustrating how sampling distributions tighten as sample size grows. n small n medium n large true mean sample mean density

The tools on the bench

Ideas that pay rent.

Central Limit Theorem · Statistical foundation
sample means → normal as n grows · SE = SD / √n
With n ≥ 30, the sampling distribution is normal regardless of source shape.
Confidence Intervals · Inferential statistics
point estimate ± margin of error · MoE = 2 × SE for 95%
A 95% CI means 95% of such intervals (over many samples) would contain the true parameter — not that there is a 95% probability this one does.
Standard Error · Sampling theory
SD of the sample mean · shrinks with √n
To halve the CI width, quadruple the sample.

How to apply

Reporting an uncertain number.

  1. Compute the point estimate and SE. Mean, SD, n. Divide.
  2. Report the range, not the number. "27,412 ± 1,750 bikes" tells the reader what you actually know.
  3. If the range is too wide, collect more data. There is no other lever.

Key reading · OpenIntro Statistics · Chapter 4

Foundations for inference.

The standard error is the most important number in practical statistics — more important than the mean. It is what lets you move from "here is a number" to "here is a range we can defend."

Report the range. Always.

← back to DA ··· m.02 A/B Testing & Hypothesis Logic →