Plotting the co-ordinates of the compositional mean on a ternary plot with ggtern

Question

I have a compositional set of data investigating sedentary behaviour, sleep and physical activity. Each of these data points is a proportion of time. I want to plot this on a ternary plot using ggtern and have done as follows which results in the below output.

library(ggtern)
library(gtools)
library(zCompositions)

comp_df <- as.data.frame(data[,c("id","PA", "SB", "Sleep")])

ggtern(data = comp_df, aes(Sleep, SB, PA)) +
  geom_mask() +
  geom_point() +
  theme_bw() + 
  theme_showarrows() +
  geom_confidence_tern(colour = "black", breaks = c(0.25, 0.5, 0.75)) 

Now this is most of the way there. However I want to add a data point that plots the compositional mean of this dataset on the ternary plot, and also labels the co-ordinates of this mean in a specific way. I've use the epicoda package from GitHub to identify the compositional mean and add the appropriate labels as below.

comp_labels <- c("Sleep", "SB", "PA")
library(epicoda)
cm <- comp_mean(comp_df, comp_labels = comp_labels, units = "hr/day")

Then, I can add this easily to my ternary plot with the addition of another geom_point() argument which results in the same output but with the additional point for the compositional mean.

ggtern(data = comp_df, aes(Sleep, SB, PA)) +
  geom_mask() +
  geom_point() +
  geom_point(data = cm) +
  theme_bw() + 
  theme_showarrows() +
  geom_confidence_tern(colour = "black", breaks = c(0.25, 0.5, 0.75)) 

What I want to be able to do is add lines to label the mean with both percentages and time like this image. Any assistance here would be very much appreciated.

Answer 1

Unfortunately you did not provide any data, so this code comes untested, but geom_[TLR]line should help you. For example to mark the mixture 20%/20%/60% you could use:

ggtern(data = comp_df, aes(Sleep, SB, PA)) +
  geom_mask() +
  geom_point() +
  theme_bw() + 
  theme_showarrows() +
  geom_confidence_tern(colour = "black", breaks = c(0.25, 0.5, 0.75)) +
  geom_Tline(Tintercept = .2) +
  geom_Lline(Lintercept = .2) +
  geom_Rline(Rintercept = .6)

That is, something like:

geom_Tline(Tintercept = cm$SB) +
  geom_Lline(Lintercept = cm$Sleep) +
  geom_Rline(Rintercept = cm$PA)

should do the trick.