All frameworks and cheat sheets

Comparison

Proportionate vs disproportionate stratified sampling

Comparison of proportionate and disproportionate stratified sampling: mirror population shares for totals, or oversample small groups for subgroup reads.
Pick the allocation by the question: the whole population, or a small group you need to read well.
Download PNGRead the full guide

A side-by-side comparison of the two ways to allocate a stratified sample. Proportionate allocation gives each stratum its population share and the best whole-population estimate. Disproportionate allocation oversamples small strata to get precise subgroup estimates, and needs weighting before you report totals. In Koji, screener questions sort each participant into a segment before the interview starts.

Where this appears

What's in this comparison

Proportionate

  • Each stratum gets the share it holds in the population
  • The sample is a miniature of the whole
  • Best estimate for the total population
  • Use it when the overall answer matters most

Disproportionate

  • You oversample small strata on purpose
  • Precise reads on small subgroups
  • Costs some whole-population accuracy
  • Weight the results before you report totals

Use this image

Free to use in your docs, decks and posts under CC BY 4.0, as long as you credit Koji with a link to this page. To embed it, paste this into your page:

html
<a href="https://www.koji.so/visuals/proportionate-vs-disproportionate-stratified-sampling"><img src="https://www.koji.so/visuals/proportionate-vs-disproportionate-stratified-sampling/proportionate-vs-disproportionate-stratified-sampling.png" alt="Comparison of proportionate and disproportionate stratified sampling: mirror population shares for totals, or oversample small groups for subgroup reads." width="1200" height="641" /></a>
<p>Source: <a href="https://www.koji.so/visuals/proportionate-vs-disproportionate-stratified-sampling">Proportionate vs disproportionate stratified sampling</a>, Koji</p>

Related

Process showing six steps for stratified sampling, from choosing a stratifying variable and defining strata to random draws and weighting.

How to run a stratified sample in six steps

Comparison of a digital shelf scrape and a real shopper across login, location, timing, search phrasing, how much of the grid they see and intent.

What a digital shelf scrape sees vs what the shopper sees

Comparison of the jingle fallacy, two different things sharing a name, and the jangle fallacy, one thing with two names, with a product example of each.

Jingle vs jangle: two naming fallacies in measurement