top of page
Search

Sampling in behavioral science

Writer: Mohsen Rafiei
Mohsen Rafiei
6 days ago
2 min read

Sampling is one of the most critical steps in any behavioural science study including UX ones, and it is usually the one we spend the least time on.



Here is why it matters so much. Your sample decides who your results are actually about. You can have a clean design, validated measures, and a careful analysis, and still end up answering a question about the wrong people. And unlike a weak analysis, a weak sample cannot be fixed after the data is in. No statistical model can tell you about people who never had a chance to be in your study.



The methods fall into two families. In probability sampling, every person in the population has a known chance of being selected. Simple random sampling draws people by chance from a full list. Systematic sampling takes every kth person on that list. Stratified sampling splits the population into subgroups and draws randomly within each one, so no group gets missed. Cluster sampling picks whole groups, like schools or clinics, instead of individuals. These methods let you generalize to a population, but they need a list of that population and a real budget.



In non-probability sampling, selection is not random. Convenience sampling takes whoever is easy to reach, which covers most participant pools and online panels. Quota sampling sets targets for subgroups and fills them with whoever shows up first, so it looks like stratified sampling without the random draw. Purposive sampling hand picks people who have the characteristics the question needs. Snowball sampling asks participants to bring in others, which helps with groups that are hard to find.



Most behavioural research runs on the second family, and that is fine. The problem starts when a convenience sample gets described as if it were a random one. Two habits help. Pick the method based on the claim you want to make, not on what is easiest. And remember that random assignment and random sampling are different things. Random assignment tells you the effect is real in your sample. Only the sampling tells you who else it applies to.


 
 
 

Comments


  • LinkedIn

©2020 by PUXLab.

bottom of page