There appear to be a large number of rules of thumb for histogram bin size and kernel selection for density plots. Are histograms and/or density plots really the best visualization for a single continuous variable? Are there other desirable options? Are there any settled best practices for visualizing continuous data that has no known distribution? Is there a good citable comprehensive reference source in the methods available and their advantages/disadvantages?

  • 2
    $\begingroup$ Q-Q plots can be quite useful, but they take some time to learn to 'read' them well. KDE's are not the only choice for density plots (e.g. there's log-spline density estimation). $\endgroup$
    – Glen_b
    Jul 27 '14 at 17:31
  • 2
    $\begingroup$ Depends on the sample size too. For really small samples, a dotplot is really probably the best, because densities, bin frequencies, quantiles (or hinges), etc. are just not knowable. $\endgroup$
    – Russ Lenth
    Jul 27 '14 at 20:05

You ask for a citable reference, I would go for one of William Cleveland's book Visualizing Data, also see his author page. See also this earlier questions: https://www.amazon.com/William-S.-Cleveland/e/B000AP9IPS/ref=dp_byline_cont_book_1 and What's a good book or reference for data visualization?.


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.