Open larsyencken opened 3 months ago
/cc @lucasrodes
Here's a good example of a benefit of keeping projections in different indicators, at least as one of our options: https://ourworldindata.org/grapher/comparison-of-world-population-projections
Another related issue we have is that currently, there is a visible gap between when the observed data ends and when the projections begin.
The problem of upper and lower confidence intervals is also a bit related (two more time series that are closely related but lead to some odd behaviour e.g. in facetting).
Probably it makes sense to keep these as different variables as they can then also have different sources (e.g. the historic data might come from a different institute than the projection?).
The relationship of different variables/indicators to each other is something that Grapher is very bad at. If it were better with this then this could also have big benefits for doing drillup/-down.
Background
Sometimes our data sources include projections into the future. Grapher handles this by using two indicators, one for the "stem" of the chart containing actual measurements, and one for the projections. The projection indicator has
isProjection: true
set, causing it to be displayed as a dotted line.Problem
Our approach makes all lines with projections into multi-indicator charts. But when charts are both multi-indicator and multi-entity, then every line needs a double-label, which is super ugly.
Impact
This tends to impact only a few charts, but quite important ones.
We have workarounds, but they require us to either
Possible approaches
"projectionAfter": 2023
)