I'm looking for a log (base 10) colour scale to colour a SOM U-matrix plot in R. Specifically, looking for a colorRampPallette that will have more bins at the low end of the distribution than the high when colour coding in original units.
I found a possible lead here https://stat.ethz.ch/pipermail/r-help/2006-July/110187.html but this solution seemed to be overly complicated.
I'm not sure where to start with this but suspect somebody has already solved this log-scale palette problem?
in ggplot one can use scale_color_gradientn. Here is an example with cars data.
ggplot(cars)+
geom_point(aes(x = speed, y =dist, color = dist))+
scale_color_gradientn(colors = colorRampPalette(colors = c("blue", "white"))(nrow(cars)),
values = scales::rescale(log(sort(cars$dist))))
To summarize, one can define a linear gradient of any number of colors with colorRampPalette function, and in scale_color_gradientn you can map any of the colors to a certain value - the spread of the colors can be linear, log or arbitrary. Since values argument accepts 0 - 1 range, scales::rescale was used on the log transformed values.
to compare, here is without transformation
ggplot(cars)+
geom_point(aes(x = speed, y =dist, color = dist))+
scale_color_gradientn(colors = colorRampPalette(colors = c("blue", "white"))(nrow(cars)), values = scales::rescale(sort(cars$dist)))
