Time series prediction: visualising path uncertainty region I am predicting a time series' future evolution and am evaluating the path uncertainty using bootstrapping.
Is there a good way to visualise the uncertainty that goes beyond simply plotting a pair of confidence bands, e.g. 90% lines? I thought of something that gives a better feeling for the distribution at each future time point: e.g. something like a contour plot or a plot where the level of transparency indicates the likelihood.
Also, is there a convenient way to compare two path predictions from different models visually? The only thing that comes to mind is two overlaid contour plots.
 A: Expanding on a comment:
My preference for showing confidence intervals is to use coloured contour plots, going from a deep coloured region for high certainty to pale colours for the edge of the distribution.  That way, you not only show the different boundaries of the confidence regions, but you also give the visual impression of a "fuzzy" region, fading from the best-prediction line to the edge of possibility.
I haven't been doing much advanced R graphics lately, so I can't help you too much with implementation, but I put together a mock-up in Javascript/D3:  http://fiddle.jshell.net/WDc3H/1/

Because I've implemented the fading colours through transparency, you can even add in a partially-overlapping second prediction+CI set and compare the two.  How easy it is to compare will depend on how much overlap you have.  You can use the checkboxes on the above link to toggle a second series, but there is really too much overlap, and the contour lines from the different confidence intervals are more dominant than the difference in colour.

It's more effective when there is greater divergence in the distributions, as in this modified example where I've extended the lines over a wider domain:
http://fiddle.jshell.net/WDc3H/2/

