Dynamic displays

A standard scatterplot is a static display that can be presented on paper as effectively as on a computer screen. In contrast, our next display for data with three variables is a dynamic display that can only be used on a computer — the information in the display is conveyed by movement.

Scatterplots can be extended to 3 dimensions

To display a data set with three numerical variables, imagine a clear plastic cube with axes on its three dimensions (height, width and depth) for the three variables. The three axes could be used to locate a small sphere in the plastic for each individual. The plastic cube would be a 3-dimensional scatterplot.

Although a computer screen is only 2-dimensional, it can display a projection of such a 3-dimensional scatterplot. Rotating such a display with the mouse gives a good feel for the shape of the underlying 3-dimensional scatter of points.

Dow Jones and NASDAQ share price indices

The diagram below shows a scatterplot (i.e. time series plot) of the Dow Jones share index on all trading days between 15 January and 17 April 2002. (15 January is labeled day 1.)

Click the checkbox Join Crosses to join successive values with lines. (This shows the trend better.)

Position the mouse in centre of the scatterplot — depending on your software, the pointer may change into a hand or there may be a target pink circle. Then drag with the mouse towards the bottom right of the screen. The display is 3-dimensional and also shows the NASDAQ share index for each day.

(It may help to imagine that the mouse is holding the surface of a sphere that surrounds the scatterplot and is using that to do the rotation.)

You may click the buttons under the display to change to one of 4 pre-set rotations. The Spin button starts the display continuously rotating.

After a little rotation, you should get a good feel for how both share prices changed over this period.

Three-dimensional scatterplots are an interesting (and occasionally useful) way to display data. They are however much overrated as an analysis technique and simpler displays are usually more effective for extracting information from multivariate data.