---
tag: PNMT15BB91
kind: concept
name: Stratified sampling
msc: [62D05]
sources:
  - url: https://en.wikipedia.org/wiki/Stratified_sampling
same_as:
  wikidata: Q1517706
  wikipedia: Stratified sampling
---

## Summary

Stratified sampling divides a finite population into disjoint groups, the strata, and draws a separate sample from each stratum.
Population quantities are then estimated by combining the stratum estimates with weights equal to the stratum shares.
When the strata are internally homogeneous, this gives more precise estimates than a [[simple random sample|Y0PTCJBPEA]] of the same total size.

## Explanations

### L1 human

To estimate the average income in a country, one can first split the country into regions and then survey some people in each region.
This guarantees that every region is represented, which a purely random draw might fail to do by bad luck.
The regional averages are then combined, each weighted by the share of the population living there.
If people within a region are similar, the combined estimate is much more accurate.