Let's say I ask 1,000 people a survey with 10 questions.

Of these 1,000 people, 50% were men and 50% were women

The number of responses, however, were as follows:

  • Men: 100 (66.6% of responses)
  • Women: 50 (33.3% of responses)

Therefore my survey responses are skewed towards men.

Should I over-sample / under-sample the responses to match the population distribution before analysing the responses?

p.s. using gender for illustrative purposes, these groups could be anything


1 Answer 1


As I understand the question, the sample has already been taken but you have the option to add to the sample. Generally speaking, a larger sample is better so there would likely be benefits for adding to the sample. However, this probably isn't necessary in your case. Instead of adding to the sample, you can poststratify on gender. This will assign greater weight to female respondents such that they are equally represented.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.