I have a plot made of many samples, usually the sampling rate is 8 samples per seconds but it may be more or less. Each and every sample has a timestamp of course.

I have a UI in which I display the plot to the user. The user may asks for 10 years of data and I'm showing the complete plot on screen. Since the pixel width of the screen is limited (let's say 1000 pixels) the plot is just a rough representation of the real one.

This means that, on a very zoomed out plot, each pixel column represents more than 1 sample, it may represents a week worth of data with thousand and thousand of samples in it.

If the user point selects on a very zoomed out plot, which sample should be "selected"? Is it worth telling the user that he/she actually selected a timerange (i.e. the pixel column, accounting for a week of data) instead of a precise point in time? Is it best to just pick one sample and show its timestamp? What if the user selects a region? (i.e. a starting point and an end point on the plot)

  • I've read your question multiple times, but I feel like it's quite difficult to fully understand your issue. Could you post a few screenshots to clarify everything? Thanks! – Tudor Teisanu Jun 8 '17 at 12:56

A graph is effective at summarizing data so that a user can more intuitively observe trends, compare related values, and discover other qualities about the underlying data points.

Knowing this, I wouldn't stress so much about "which of the thousands of data points under this pixel should show on hover?" I would, however, provide the user with the ability to zoom so they can get as granular as they'd like.

The following image summarizes what I would consider to be a well-designed graph (courtesy of University of Pittsburg).

Graph with user interaction

  • ok.. but as you can see you are selecting a point on the plot albeit that point is not a single value, which value should be picked instead? the first one of the group? the middle one? or maybe is it irrelevant? – Gianluca Ghettini Jun 8 '17 at 14:38
  • @GianlucaGhettini Personally, I think it's unimportant, because if the user needs that amount of precision, I don't think a graph is the best method to disseminate that data. They'd be better with a table or report of some kind. If you're still looking for an answer, I'd say just pick the first. I would probably advise against showing an average unless you disclose that it's an average so you don't falsify your data. – maxathousand Jun 8 '17 at 14:49

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.