Context: I have a survey where I have deliberately oversampled particular groups of respondents. I am aggregating these groups and therefore I am using measures such as the weighted mean to ensure my results are representative of the population.

However, in circumstances where I want to show particular observations or produce graphs such as histograms or scatter plots, how do I best approach it? Is it simply a case of plotting the raw unweighted data? Is it generally accepted that reports on surveys of this kind will jump between raw unweighted data and weighted summary statistics or is it recommended to have clear divides from when I am talking about weighted numbers and when I am not?

  • $\begingroup$ I think that you should make it very clear when you use an aggregate (weighted in your case) measurement and not. I don't know what kind of survey you are doing, but in general I would make a clear distintion between both. $\endgroup$ – chuse Jan 26 '15 at 14:55
  • $\begingroup$ the R survey package has a svyplot command that's useful for visualizing weighted data points. $\endgroup$ – Anthony Damico Jan 27 '15 at 4:45

My understanding is if data is weighted then you should only ever report/chart using the weighted results but use unweighted base sizes.

| cite | improve this answer | |
  • $\begingroup$ Thanks for responding. When you say 'weighted results' I suspect you are referring to a distinct calculation being performed. What about in those situations where you simply want to visualise raw data? $\endgroup$ – NickB2014 Jan 26 '15 at 20:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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