calculate color for higher level rectangles from weighted average of lower level when type='value'

[ellisp]

Consider the example below, converting a {treemap} object for use by d3tree2. When you click down to the level where the value that has been given to vColor exists, Borduria looks like this:

The average value/color intuitively should be 2.6, hence a pale yellow/green, but instead we get purple/10, more or less aribtrarily (I'm not sure exactly why, but could imagine reasons).

It would be very useful if in this case the color assigned to the higher level rectangles was a weighted average of the values assigned to color the level below, until you reach the rectangles that have values actually assigned from originally. (I might have a crack at this myself down the track, but not in the immediate next few weeks.)

library(treemap)
library(d3treeR)
library(dplyr)

mock <- data.frame(
   Country = rep(c("Ruritania", "Borduria"), 3),
   Industry = rep(c("Accommodation", "Manufacturing", "Flower growing"), each = 2),
   GDP = c(1000, 500, 20, 2000, 1800, 70), 
   Growth = c(-5, 4, -3, 2, -7, 10)
   )

tm <- treemap(mock, index = c("Country", "Industry"), vSize = "GDP", vColor = "Growth", 
               type = "value", palette = "Spectral")
d3tree2(tm, rootname = "World")

# what should the country colours be? Probably mean vColor value, weighted by vSize:
mock %>%
   group_by(Country) %>%
   summarise(AvGrowth = round(sum(GDP * Growth) / sum(GDP), 1))

[ignacio82]

Try adding fun.aggregate:

tm <- treemap(mock, index = c("Country", "Industry"), vSize = "GDP", vColor = "Growth", 
              type = "value", palette = "Spectral", 
              fun.aggregate="weighted.mean")
d3tree2(tm, rootname = "World")

[timelyportfolio]

Thanks @ignacio82. @ellisp treemap was not originally designed for this since colors only appear at the lowest level. Fortunately @mtennekes quickly added in mtennekes/treemap#18. Would love feedback.

[ellisp]

Perfect, exactly what I was looking for, thanks all.