17

I would like to plot some different data items using ggplot2, using two different colour scales (one continuous and one discrete from two different df's). I can plot either exactly how I would like individually, but I cannot make them work together. It seems like you cannot have two different colour scales operating in the same plot? I have seen similar questions here and here, and this had led me to believe that what I would like to achieve is simply not possible in ggplot2, but just in case I am wrong I would like to illustrate my problem to see if there is a work-around.

I have some GIS stream data which has some categorical attributes attached to it, which I can plot (p1 in code below) to get: enter image description here

I also have a set of locations which have a continuous response, which I can also plot (p2 in code below) to get: enter image description here However I can't combine the two (p3 in code below). I get this error

Error in scales[[prev_aes]] : attempt to select less than one element

Commenting out the line scale_colour_hue("Strahler order") + changes the error to

Error: Discrete value supplied to continuous scale

Basically it seems that ggplot2 uses the same scale type (continuous or discrete) for the geom_path call and the geom_point calls. So when I pass the discrete variable, factor(Strahler), to the scale_colour_gradientn scale, the plot fails.

Is there a way around this? It would be amazing if there was a data argument to a scales function to tell it where it should be mapping or setting attributes from. Is this even possible?

Many thanks and reproducible code below:

library(ggplot2) ### Download df's ### oldwd <- getwd(); tmp <- tempdir(); setwd(tmp) url <- "http://dl.dropbox.com/u/44829974/Data.zip" f <- paste(tmp,"\\tmp.zip",sep="") download.file(url,f) unzip(f) ### Read in data ### riv_df <- read.table("riv_df.csv", sep=",",h=T) afr_df <- read.table("afr_df.csv", sep=",",h=T) vil_df <- read.table("vil_df.csv", sep=",",h=T) ### Min and max for plot area ### xmin <- -18; xmax <- 3; ymin <- 4; ymax <- 15 ### Plot river data ### p1 <- ggplot(riv_df, aes(long, lat)) + geom_map( mapping = aes( long , lat , map_id = id ) , fill = "white" , data = afr_df , map = afr_df ) + geom_path( colour = "grey95" , mapping = aes( long , lat , group = group , size = 1 ) , data = afr_df ) + geom_path( aes( group = id , alpha = I(Strahler/6) , colour = factor(Strahler) , size = Strahler/6 ) ) + scale_alpha( guide = "none" ) + scale_colour_hue("Strahler order") + scale_x_continuous( limits = c( xmin , xmax ) , expand = c( 0 , 0 ) ) + scale_y_continuous( limits = c( ymin , ymax ) , expand = c( 0 , 0 ) ) + coord_map() print(p1) # This may take a little while depending on computer speed... ### Plot response data ### p2 <- ggplot( NULL ) + geom_point( aes( X , Y , colour = Z) , size = 2 , shape = 19 , data = vil_df ) + scale_colour_gradientn( colours = rev(heat.colors(25)) , guide="colourbar" ) + coord_equal() print(p2) ### Plot both together ### p3 <- ggplot(riv_df, aes(long, lat)) + geom_map( mapping = aes( long , lat , map_id = id ) , fill = "white" , data = afr_df , map = afr_df ) + geom_path( colour = "grey95" , mapping = aes( long , lat , group = group , size = 1 ) , data = afr_df ) + geom_path( aes( group = id , alpha = I(Strahler/6) , colour = factor(Strahler) , size = Strahler/6 ) ) + scale_colour_hue("Strahler order") + scale_alpha( guide = "none" ) + scale_x_continuous( limits = c( xmin , xmax ) , expand = c( 0 , 0 ) ) + scale_y_continuous( limits = c( ymin , ymax ) , expand = c( 0 , 0 ) ) + geom_point( aes( X , Y , colour = Z) , size = 2 , shape = 19 , data = vil_df ) + scale_colour_gradientn( colours = rev(heat.colors(25)) , guide="colourbar" ) + coord_map() print(p3) #Error in scales[[prev_aes]] : attempt to select less than one element ### Clear-up downloaded files ### unlink(tmp,recursive=T) setwd(oldwd)

Cheers,

Simon

Improve this question
5
  • 1
    The problem isn't so complicated as you might think. In general, you can only map an aesthetic once. So calling scale_colour_* twice makes no sense to ggplot2. It will try to force one into the other. Commented Jul 16, 2012 at 17:01
  • @joran So there is currently no way to map one colour aesthetic from a dataframe to a continuous scale and another colour aesthetic from a different dataframe to a discrete scale? It would be handy, would it not, if it were possible to specify data to scales? Commented Jul 16, 2012 at 17:08
  • 1
    You can't have multiple colour scales in the same graph, regardless of whether either one is continuous or discrete. The package author has said that they have no intention of adding this, either. It is rather complicated to implement and would make it too easy to make incredibly confusing graphs. (Multiple y axes will never be implemented for similar reasons.) Commented Jul 16, 2012 at 17:47
  • @joran Thanks for clearing that up. Whilst I understand the caution behind not being able to create multiple colour scales (or other aesthetics), I would prefer it if the onus was on content creators (and the peer-review process) to allow meaningful and clear plots, rather than being deliberately limited by the tools. However, I quite understand not implementing this for reasons of complexity. ggplot2 already has such an extensive feature set I imagine this would be highly difficult to integrate. I do absolutely love aesthetic mapping! Commented Jul 16, 2012 at 18:10
  • @joran Can you add your comments as an answer, so this question can be marked as answered? Commented Oct 19, 2012 at 8:47

4 Answers 4

Reset to default
14

You can do this. You need to tell grid graphics to overlay one plot on top of the other. You have to get margins and spacing etc, exactly right, and you have to think about the transparency of the top layers. In short... it's not worth it. As well as possibly making the plot confusing.

However, I thought some people might like a pointer on how to acheive this. N.B. I used code from this gist to make the elements in the top plot transparent so they don't opaque the elements below:

grid.newpage() pushViewport( viewport( layout = grid.layout( 1 , 1 , widths = unit( 1 , "npc" ) ) ) ) print( p1 + theme(legend.position="none") , vp = viewport( layout.pos.row = 1 , layout.pos.col = 1 ) ) print( p2 + theme(legend.position="none") , vp = viewport( layout.pos.row = 1 , layout.pos.col = 1 ) )

See my answer here for how to add legends into another position on the grid layout.

enter image description here

Improve this answer
Sign up to request clarification or add additional context in comments.

Comments

5

The problem isn't so complicated as you might think. In general, you can only map an aesthetic once. So calling scale_colour_* twice makes no sense to ggplot2. It will try to force one into the other.

You can't have multiple colour scales in the same graph, regardless of whether either one is continuous or discrete. The package author has said that they have no intention of adding this, either. It is rather complicated to implement and would make it too easy to make incredibly confusing graphs. (Multiple y axes will never be implemented for similar reasons.)

Improve this answer

Comments

3

I don't have the time at the moment to provide a complete working example, but there's another way to do this that deserves to be mentioned here: Fill and border colour in geom_point (scale_colour_manual) in ggplot

Basically, using geom_point in conjunction with shape = 21, color = NA allows you to control the color of a series of points using the fill rather than color aesthetic. Here's what my code looked like for this. I understand that there's no data provided, but hopefully it provides you with a starting point: 

biloxi + geom_point(data = filter(train, primary != 'na'), aes(y = GEO_LATITUDE, x = GEO_LONGITUDE, fill = primary), shape = 21, color = NA, size = 1) + scale_fill_manual(values = c('dodgerblue', 'firebrick')) + geom_point(data = test_map_frame, aes(y = GEO_LATITUDE, x = GEO_LONGITUDE, color = var_score), alpha = 1, size = 1) + scale_color_gradient2(low = 'dodgerblue4', high = 'firebrick4', mid = 'white', midpoint = mean(test_map_frame$var_score))

Notice how each call to geom_point invokes a different aesthetic (color or fill)

Improve this answer

Comments

3

With Elio Campitelli's ggnewscale it is now much simpler to combine different color scales (even discrete and continuous) in the same ggplot:

# OP data is no longer available, # example by Elio Campitelli from https://eliocamp.github.io/ggnewscale/, modified library(ggplot2) library(ggnewscale) # "melt(volcano)" topography <- data.frame(expand.grid(x = 1:nrow(volcano), y = 1:ncol(volcano)), z = as.vector(volcano)) # point measurements of something at a few locations set.seed(42) measurements <- data.frame(x = runif(30, 1, 80), y = runif(30, 1, 60), thing = sample(letters[1:5], 30, replace=TRUE) ) ggplot(mapping = aes(x, y)) + geom_contour(data = topography, aes(z = z, color = after_stat(level))) + # Color scale for topography scale_color_viridis_c(option = "D") + # geoms below will use another color scale new_scale_color() + geom_point(data = measurements, size = 3, aes(color = thing)) + # Color scale applied to geoms added after new_scale_color() scale_color_viridis_d(option = "A")

Keep in mind that multiple color scales in the same plot can be confusing and that there are dedicated packages for plotting maps with ggplot (like ggmap and sf).

Improve this answer

Comments

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.