Sometimes I need to subsample a dataset for my analysis/modelling. What are the strategies/things to check for to select a subsample that's representative of the entire sample? And in addition, how to evaluate the 'representativeness' of the subset taken?

I visited this post, but I think its focus is on reducing the volume of the data instead of getting a qualitatively representative sample of the larger set. Besides, I'm asking for non-time series data, which is something more general.

I also visited this post, but find it not that relevant.

It would be nice to hear how things are done in (industrial) practice. TIA.

  • 1
    $\begingroup$ As mentioned in the first post for non-sequential data, have you tried simply sampling random rows from the dataset? $\endgroup$
    – Oxbowerce
    Commented Aug 18, 2022 at 10:01
  • $\begingroup$ @Oxbowerce, thanks for the recommendation. I've edited the question to reflect what I truly want to ask for, in addition to how to sample, also how to evaluate. Thanks for your inputs :) $\endgroup$
    – Student
    Commented Aug 22, 2022 at 1:19


Your Answer

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

Browse other questions tagged or ask your own question.