legend updates needed for divergent colormaps for continuous overlay variables

[danicahelb]

From @bobular

In GEMS1, BMI-for-age z-score ranges from -15 to +273, but when i use this term as an overlay, the legend indicates it ranges from -30 to +30. (I thought it could have something to do with not having data for the participant whose BMI-for-age z-score was 273 but checked and this is not the case)

I've had a quick look. The code is creating a symmetrical gradient around zero and is using the curated displayRangeMin and displayRangeMax in preference to the data-derived rangeMin and rangeMax. (This preference has been standard practice until now, I believe.) So this is why we get -30 to 30 (because it creates a symmetric range using max(abs())

[danicahelb]

In 1/25/2023 data viz group, we discussed how to set the default range for gradient overlays and how does it map to hues in the gradient

decision is that the gradient colormap should be consistent w what we do for the axes:

“Axes representing user-selected continuous variables will scale according to study metadata (meaning use the annotated display range from the owl file when available, expand to the full data set range if we see any values in the response which exceed that)”

Unlike what we do for the axes, if the resulting range does not include 0 we should NOT expand the range to do so. currently we do force the gradient map to run through 0 even when the data doesn't and this makes it hard to distinguish the colors. see this example from GEMS1:

[danicahelb]

also, in the 1/25/2023 dataviz meeting, we decided that -- for diverging colormaps -- when the range isn't symmetrical (ie, -5 to 25 instead of -25 to 25), we want to cut the gradient off on the early side instead of forcing the hues to be mapped to the absolute value of the larger

[moontrip]

Hi @danicahelb May I ask two questions to clarify?

First, for the case of Bob's screenshot above, displayMinMax = (-30, 30) and rangeMinMax = (-15.3, 273.91), then should the new range be (-30, 273.91)? Second, do we need to consider subsetted (filtered) data? For example, assuming that the range of the displayMinMax and rangeMinMax are the same with the first question and somehow subsetted data has MinMax = (10, 20), then do we need to respect the combination of the displayMinMax and rangeMinMax (e.g., -30, 273.91) or just use data-based MinMax (10, 20) for the colormap range?

In 1/25/2023 data viz group, we discussed how to set the default range for gradient overlays and how does it map to hues in the gradient

decision is that the gradient colormap should be consistent w what we do for the axes:

“Axes representing user-selected continuous variables will scale according to study metadata (meaning use the annotated display range from the owl file when available, expand to the full data set range if we see any values in the response which exceed that)”

Unlike what we do for the axes, if the resulting range does not include 0 we should NOT expand the range to do so. currently we do force the gradient map to run through 0 even when the data doesn't and this makes it hard to distinguish the colors.

[bobular]

This seems complicated. I can't see a clear feature request in my original message.

Can we clarify what we mean here:

for diverging colormaps -- when the range isn't symmetrical (ie, -5 to 25 instead of -25 to 25), we want to cut the gradient off on the early side

So let's say a regular -25 to +25 color map goes from dark blue to dark red.

Does the legend for -5 to +25 go from very light blue to dark red?

[bobular]

DK's question about filtered ranges might depend on whether the overlay variable was directly or indirectly filtered - aka "filter-aware" behaviour?

[moontrip]

@bobular First of all, thank you for looking into this ticket 👍 Concerning your question, all legends (both normal and colormap) utilize pre-defined color scale from the first to the end. Thus, if the range will be -5 to +25, I think it should be from dark blue to dark red which would be more consistent throughout the style of colormap: that is, a minimum is dark blue and a maximum is dark red.

[moontrip]

DK's question about filtered ranges might depend on whether the overlay variable was directly or indirectly filtered - aka "filter-aware" behaviour?

@bobular Overlay variable contains metadata for rangeMin/Max and/or displayMin/Max. I have checked those parameters after subsetting/filtering, but they did not change. Actually this is the reason why we computed data-based min/max for Scatterplot, Lineplot, etc. (for axis range controls)

Can you please confirm if you intended to consider filtered range as well? I think it should be like other Vizs (Scatterplot etc.) but I wanted to make sure of it before going for it because while it would be doable, it will take some time. I have already made a change to handle displayMin/Max and rangeMin/Max but data-based Min/Max in this case is somewhat tricky because we need to firstly compute data-based Min/Max (including facet case) and then change data based on the Min/Max (marker colors, etc.): on the other hand, for an example of Scatterplot Viz, we just used data-based Min/Max for setting axis range control, not changing data itself.

[bobular]

@bobular First of all, thank you for looking into this ticket +1 Concerning your question, all legends (both normal and colormap) utilize pre-defined color scale from the first to the end. Thus, if the range will be -5 to +25, I think it should be from dark blue to dark red which would be more consistent throughout the style of colormap: that is, a minimum is dark blue and a maximum is dark red.

Oh, I think it needs to use the middle (white-ish) color for zero. @danicahelb please could you confirm this and my question above.