Regular readers of Webtorque will recall that I put forward a theory of statistical information some months ago, which probably needed to be read in the style of the Monty Python sketch of a similar vein.

Today, I have another theory about the visual presentation of statistical information, and it is a theory that is this:

The value of a statistic decreases exponentially to the amount of non-statistical information included with it.

This is therefore a theory of chart junk: if you draw a graph, and show the X and Y axes as being made out of clocks and cherries respectively, you have decreased the value of your chart by an amount corresponding to the two distracting things you’ve added to it. The same is true of diagrams in general. I say it’s exponential, but if it’s not then it’s certainly not a linear function.

The corollary (oh yes) of this is that it’s pretty hard to do much damage to a chart by removing things, so they’re usually good candidates for reduction.