d3-foresight
d3-foresight copied to clipboard
reichlab
Make legend colors more sensible
^ this is a mess. Showing around 22 colors is already not that useful in the main plot area. But here, because of the randomization, the legend also becomes useless. The problem is that (other than a little bit of distinction) the colors are very confusing and one can't figure out what to make of them.
A possible solution might be to sort the preset colors in an order based on the deriving palette and skip colors which are too similar. This might at least let us create a mental map of models (since models are sorted by team names) to general colors.
Right now, we have same colors sorted by hue
This looks like an improvement but since there are 22 colors used out of 30, we can remove some of them based on how similar they are to neighbours.
Another issue is that a systematic ordering of colors has made the later models more visible since their curves are drawn at the end. For example, you can see a lot of blue/green in the plots at flusightnetwork.io
Could we have a single color represent each "team" (based on unique model name prefix) and then within a team have different hues? I.e. Delphi has hues of green, ReichLab hues of blue, etc..?
In terms of the issue with overlapping, maybe we could have a default be to not show confidence intervals if there are >k models (where k=8?) or something like that?
