Stratified sampling
Stratified sampling is a probability sampling method that divides a population into subgroups, called strata, that share a characteristic, then draws a random sample from each one. Every stratum is guaranteed a place in the sample, either in proportion to its size or deliberately oversampled.
How it works
The researcher chooses the stratifying variable, such as region, age band or customer tier, and splits the sampling frame into non-overlapping strata. Each stratum is then sampled randomly. In proportionate stratification, each stratum's share of the sample matches its share of the population; in disproportionate stratification, small strata are oversampled so they can be analyzed on their own, and weights restore the overall balance.
Stratifying on a variable related to the outcome reduces sampling error compared with simple random sampling of the same size, because variation between strata is removed. It needs a frame that records the stratifying variable for everyone.
A worked example
A retailer surveys customers across 4 regions, one of which holds only 5 percent of customers. A simple random sample of 1,000 would include about 50 from it, too few to analyze. Stratifying with 250 per region, then weighting the results back to the population, gives reliable figures for every region.
Related
Automate this in Pulse
When the strata are countries, one Pulse study run across those markets at once keeps the questions identical in every stratum, one fewer reason for the strata to differ.