Edwy
AI Primer

Statistics · Module 7

Estimation, sampling, and bias

Estimation

Using a sample of data to guess a property of the whole population. "What is the average customer spend?" answered with 500 receipts.

Sampling

How you select data shapes everything that follows. Bad samples give confident but wrong answers.

Variance + bias

Variance: how much your estimate jiggles. Bias: how systematically wrong it is. Every model trades one against the other.

central-limit intuition

Averages of many independent measurements tend to become more stable than the individual measurements.

Generalisation

Whether a finding on past data still holds on tomorrow’s data. Most AI failures are generalisation failures.

Applied frame

Ask what population the sample represents, what uncertainty remains, and how the estimate could drift.

Diagram cue

The diagram shows samples becoming estimates, with bias and variance kept visible.

Sampling bias diagram showing sample frame, bias, variance, and final estimate
A model cannot repair evidence that was sampled from the wrong frame.
Estimation, sampling, and bias animated explainer
Owned Edwy animated explainer for Estimation, sampling, and bias.

Sample design shapes estimate quality before modeling begins. Key motion beats: sample frame, bias, variance, estimate.

Estimation check

0 of 1 questions completed locally.

1. A biased sample can produce a confident but wrong estimate.

Answer feedback appears here.

Reader progress is stored locally in this browser.

Source slide 8