All frameworks and cheat sheets

Guide

Five stratified sampling mistakes to avoid

List of five stratified sampling mistakes: irrelevant variables, overlapping strata, too many strata, no weights and mistaking quota sampling for it.
Most of these break the allocation before a single person is sampled.

Five common mistakes in stratified sampling and how to avoid them: choosing a stratifying variable that has nothing to do with the outcome, strata that overlap or leave people out, slicing a small sample into too many groups, reporting an oversampled design without weights, and filling strata by convenience, which turns the study into quota sampling. Check your design against it before you start recruiting.

Where this appears

What's in this guide

  1. Stratifying on an irrelevant variable: If it does not relate to the outcome, you add work and gain no precision.
  2. Overlapping or incomplete strata: Each person must fit exactly one stratum, and together the strata cover everyone.
  3. Too many strata, too few people: A small sample cut a dozen ways leaves every group too thin to analyze.
  4. Forgetting to weight: Weight an oversampled group before you report figures for the whole population.
  5. Calling quota sampling stratified: Stratified sampling picks at random inside each stratum. Filling strata with whoever is handy is quota sampling.

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/stratified-sampling-mistakes-to-avoid"><img src="https://www.koji.so/visuals/stratified-sampling-mistakes-to-avoid/stratified-sampling-mistakes-to-avoid.png" alt="List of five stratified sampling mistakes: irrelevant variables, overlapping strata, too many strata, no weights and mistaking quota sampling for it." width="1080" height="1080" /></a>
<p>Source: <a href="https://www.koji.so/visuals/stratified-sampling-mistakes-to-avoid">Five stratified sampling mistakes to avoid</a>, Koji</p>