Summary
This stems from #1311 and #1315. We currently use "normalized variance explained" as the measure of "variance explained" in our static component plots and in the donut plot. I propose we use marginal R-squared (effectively the mean R-squared between each component and the optimally combined data) instead. This measure of variance explained is useful on its own, since it doesn't depend on other components or the regression model used, although that means that values won't sum to 100.
Summary
This stems from #1311 and #1315. We currently use "normalized variance explained" as the measure of "variance explained" in our static component plots and in the donut plot. I propose we use marginal R-squared (effectively the mean R-squared between each component and the optimally combined data) instead. This measure of variance explained is useful on its own, since it doesn't depend on other components or the regression model used, although that means that values won't sum to 100.