BigTimeStats
BigTimeStats

Reputation: 447

How to pass variable into ggplot aes / ..y.. part of function

I'm looking at passing transform into the below function to be able to line up the secondary axis... however, I get an error:

Error in y * transform : non-numeric argument to binary operator

library(ggplot2)
a.samples <- rbeta(10000, 25, 75)
b.samples <- rbeta(10000, 35, 65)

df <- data.frame(x = b.samples/a.samples)

p <- ggplot(df, aes(x = x)) + geom_histogram(fill = 'grey50', alpha = .6)

transform <- max(ggplot_build(p)$data[[1]]$ymax)

p + geom_line(aes(y = ..y.. * transform), stat='ecdf') + # error here
    geom_vline(xintercept = 1, linetype = 'dashed') +
    scale_y_continuous(sec.axis = sec_axis(~./transform, name = "ecdf")) +
    guides(color = FALSE)

If I explicitly write the value 1260 into the aes argument, it works.

...geom_line(aes(y = ..y.. * 1260), stat='ecdf')...

Upvotes: 1

Views: 406

Answers (2)

Sebastian Hoyos
Sebastian Hoyos

Reputation: 176

So your above code oddly worked for me without any other errors other than the binwidths being off so I don't know what was generating the error but I just adjusted it with adding stat_ecdf()

library(ggplot2)
a.samples <- rbeta(10000, 25, 75)
b.samples <- rbeta(10000, 35, 65)

df <- data.frame(x = b.samples/a.samples)

p <- ggplot(df, aes(x = x)) + geom_histogram(fill = 'grey50', alpha = .6)

transform <- max(ggplot_build(p)$data[[1]]$ymax)
transform
p + stat_ecdf(aes(y = ..y.. * transform),geom = "line") + 
  geom_vline(xintercept = 1, linetype = 'dashed') +
  scale_y_continuous(sec.axis = sec_axis(~./transform, name = "ecdf")) +
  guides(color = FALSE)

Upvotes: 1

BigTimeStats
BigTimeStats

Reputation: 447

Was able to get this to work with aes_string

p + geom_line(aes_string(y = paste0('..y.. * ', transform)), stat='ecdf') + 
    geom_vline(xintercept = 1, linetype = 'dashed') +
    scale_y_continuous(sec.axis = sec_axis(~./transform, name = "ecdf")) +
    guides(color = FALSE)

Upvotes: 1

Related Questions