Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

This looks suspiciously like logarithmic scaling which, as XKCD readers will attest, is a pretty typical technique for representing statistics which have differences measured in orders-of-magnitude. I'm not sure what all the panic on here is for, it's not as if Nielson only bumped Nokia, RIM and Microsoft's column widths.


  > This looks suspiciously like logarithmic scaling which,
  > as XKCD readers will attest, is a pretty typical
  > technique for representing statistics which have
  > differences measured in orders-of-magnitude.
It is, but never implicitly. Sure, you can get the actual proportions from reading the markings at the bottom, but then what's the point of visualizing in the first place? Either make your logarithmic scaling explicit and clear (as is done in scientific graphs), or use generally assumed linear scaling.

  > it's not as if Nielson only bumped Nokia, RIM and
  > Microsoft's column widths.
No, but it amounts to the same thing.


Why use a logarithmic scaling for an area chart? Given the data it's presenting, a logarithmic chart is completely inappropriate. If I take the rectangle of "HTC / Android" and add it to the rectangle of "HTC / Windows", what do I get?


If it was proportional you wouldn't be able to read the windows label in the first place. I like this setup. Works as well here as it does in disk manager, the first example I can think of commonly visiting with log scaled area.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: