R-Scatterplot: How to match palette() to not incrementing color-column

Question

I want to visualize the feature space in my training areas using scatterplots. I use a nested classification schema, so my class values are not incrementing but sorted in categories like 100 (mainclass), 110 (subclass)... I have color values already specified for each class by the GIS I'm using to digitalize the training areas, so I want to have them in the scatterplot.

I already tried to update the palette with my color vector but the results are not satisfying at all. After some experimentation I noticed that for incrementing class values the result would be satisfying.

Here is a representative example:

b1 <- c(1:50)
b2 <- runif(b1, b1-10, b1+10)
val1 <- c(rep(1,10), rep(2,10), rep(3,10), rep(4,10), rep(5,10))
val2 <- c(rep(110,10), rep(120,10), rep(200,10), rep(300,10), rep(800,10))
df <- data.frame(b1, b2, val1, val2)

par(mfrow=c(2,2))

palette("default")
plot(df$b1, df$b2, col= df$val1, pch = 20)
plot(df$b1, df$b2, col= df$val2, pch = 20)

palette(c("#FF0000", "#FB5DF6", "#AF3C96", "#857EC7", "#E8D1D1"))
plot(df$b1, df$b2, col= df$val1, pch = 20)
plot(df$b1, df$b2, col= df$val2, pch = 20)

which will result in:

The plot at bottomleft would be satisfying. How can I match my palette to the not incrementing categories on the bottomright image column?

Answer 1

You need some sort of lookup from the coding ("100","101" etc) to colours. An easy way to do this would be to convert the coding to a character value, then convert that to a "factor". Factors have an underlying numerical storage value from 1 to N, and a visible text representation.

If you do:

> df$f_val2 = as.factor(as.character(df$val2))

then plot:

> plot(df$b1, df$b2, col= df$f_val2, pch = 20)