Previous discussion was archived aton 2018-12-30.
Hi Waldir, I noticed you changed Wikidata:WikiProject sum of all paintings/Property statistics/Header to make it more accessible. The previous color scheme made by User:Wittylama was from bad (0% red), through neutral to good (100% green). The new color scheme is from empty (0% white) to full (100% blue). This changed the intention of what we wanted to show. Maybe you can change it to a accessible color scheme that goes from bad, trough neutral to good? Probably a color scheme for the company results (making a profit or loosing money) would cover this. Also @TheDJ: because he is into improving accessibility. Multichill (talk) 13:02, 21 May 2019 (UTC)
- Hey @Multichill:! I'm not sure I understand what you're asking. What you're describing is a diverging color scheme, and colorbrewer (which I used for the update I made) does have those, so we can certainly use one for cases like what you described -- indeed, such schemes are typically recommended for that kind of data, whose values can move in two opposite directions away from a central anchor point (e.g. loss vs. profit vs. break even, or height vs. depth vs. sea level, etc.).
- On the other hand, for percentages (which is what the colors are currently calculated from), there's nothing particularly remarkable about 50%; indeed the progression is linear from 0% to 100%, and for those cases a sequential color scheme is the right choice. Using a diverging color scheme for such data actually creates an artificial inflection in the visualization that distorts the perception of the underlying data.
- Are you perhaps suggesting we have two color schemes, one for linearly increasing data, and one for bimodal data? That can certainly be done, but people would have to deliberately choose which one to use (either via two templates or via a parameter in the template). How does that sound? --Waldir (talk) 19:37, 22 May 2019 (UTC)
- We chose a bimodal on purpose. Having less than 50% coverage on any of these properties means it's very incomplete and should have a "bad" color. This is intentional to encourage people. It's quite motivational getting rid of the last red block for a collection. So yes, two schemes would be best.
- By the way, at first I look at a dynamic scale based on the averages for coverage of each property and to use that instead of the hard coded 50%. I didn't implement that for the because it was too much work, but that might be a good third scheme. Multichill (talk) 21:14, 22 May 2019 (UTC)
- I see your point. Actually the color scheme does have different hues on each end of the scale, but I can see how it can appear greenish (= good-ish) throughout. While we could abandon linearity and use a diverging color map such as this one, I believe we may be able to have the cake and eat it to: in other words, a color scale that features qualitatively different hues that can be associated with "bad" and "good" ends of the spectrum, but still remains monotonically increasing in perceived lightness (which also makes it suitable for black-and-white printing, etc.). Matteo Niccoli's perceptual rainbow, aka CubeYF, seems to be a good candidate. WDYT? --Waldir (talk) 23:41, 22 May 2019 (UTC)
Can you translate and upload en:Vehicle registration plates of Azerbaijan, en:Road signs in Azerbaijan, en:Category:Road infrastructure in Azerbaijan and en:Category:Road transport in Azerbaijan in Portuguese Wikipedia?