ggplot sec_axis can I use a vector for the trans formula?

Question

I would like to label the x-axis with two non-linearly but monotonically related values. One is a decile and the other is a mean within the decile. Since there is no algebraic transformation between the two, I though I might use a vector but ggplot gives me an error message that the transformation must be monotonic. How is it checking and is there way to do what I trying?

df <- 
  data.frame(x = rnorm(500)) %>%
  mutate(y = 1 + 2 * x) %>%
  arrange(x) %>% 
  mutate(decile = ceiling(10 * 1:500 / 500)) %>% 
  group_by(decile) %>%
  summarize(x = mean(x), y = mean(y))

 sec_axis_label = df$x

df %>% 
  ggplot(aes(decile, y)) +
  scale_x_continuous(expand = c(0, 0), 
    sec.axis = sec_axis(~ sec_axis_label[.])) +  
   geom_point() 

Answer 1

The sec_axis requires a transformation (formula), not a lookup. To see what's going on, I'll change it as follows:

sec_axis_label <- function(z) { browser() ; 1 ; }
df %>% 
   ggplot(aes(decile, y)) +
   scale_x_continuous(expand = c(0, 0), 
     sec.axis = sec_axis(~ sec_axis_label(.))) +  
   geom_point() 

We can explore and see what's going on:

# Browse[2]> 
str(z)
#  num [1:1000] 1 1.01 1.02 1.03 1.04 ...
# Browse[2]> 
range(z)
# [1]  1 10
# Browse[2]>
Q

It's giving you enough potential values so that the transformation will return a meaningful range of other-axis-values so that it determine the best pretty-numbers.

So we need a transformation of sorts. If we assume interpolation is safe (I don't know if it is safe for you), we can do:

sec_axis_label <- function(a) approx(df$decile, df$x, xout=a)$y
df %>% 
  ggplot(aes(decile, y)) +
  scale_x_continuous(expand = c(0, 0), 
    sec.axis = sec_axis(~ sec_axis_label(.))) +  
   geom_point()