ggplot2 log scale of y axis causing curved lines

Question

I’d like to graph some data as means of groups over time, with lines connecting the different mean time points for each group.

The code for this is:

line<-ggplot(dat, aes(Time, Cortisol.ngmL, shape=T))

line+
stat_summary(fun.y=mean, geom="point", size=4, aes(group=T))+
stat_summary(fun.y=mean, geom="line", aes(group=T), linetype="dashed", lwd=0.7)

But…I want the y axis logged (log10). And when I do this the lines connecting the groups across time become curved (code below)

line<-ggplot(dat, aes(Time, Cortisol.ngmL, shape=T))

line+
stat_summary(fun.y=mean, geom="point", size=4, aes(group=T))+
stat_summary(fun.y=mean, geom="line", aes(group=T), linetype="dashed", lwd=0.7)+
coord_trans(y="log10")

Does anyone know a way I can have a log scale and straight lines?

We can't see your data, so without knowing what it looks like, it's hard to give a definite answer. My best guess is that the graph is correct. For a log scale, the data need to follow an exponential curve to be straight. Perhaps your data are not exponention? — Joel Carlson
It's an interesting problem - how to plot a straight line between two points on a logarithmic axis. Here's how I did it with a specific example but I'm not at all sure how to generalise that for any two points. The red line is straight on a normal scale. The blue line has been munged to look straight on the logarithmic scale. library(ggplot2); x <- seq(from=1, to=10, by=0.1); ggplot() + geom_line(aes(x = x, y = x), colour = "red") + geom_line(aes(x = (9/log10(10))*log10(x) + 1, y = x), colour = "blue") + coord_trans(y="log10") No idea how to generalise that for any two points though — Mist

Zoltán Magos Zoltán Magos · Accepted Answer · 2016-03-28T18:15:23

I use this function to connect points with straight lines in log scale:

log_line <- function(x, y, n = 1000) {
  l <- lapply(2:length(x),
              function(i, n) {
                xl <- seq(x[i - 1], x[i], (x[i] - x[i - 1]) / n)
                yl <- exp(log(y[i]) + (xl - x[i]) * (log(y[i]) - log(y[i - 1])) / (x[i] - x[i - 1]))
                return(data.frame(x = xl, y = yl))
              },
              n)

  return(do.call(rbind, l))
}

The arguments are the x and y coordinates of the points you want to connect with a straight line in log scale and the n number of points you want to predict between each pair of original points.

This function fits a straight line in log scale between each points, predicts n new points between the two original points and than converts them back to the original scale. The output is a data frame with the predicted values x and y coordinates.

It can be easily added to a ggplot:

v1 <- 1:10
v2 <- exp(-v1)

ggplot() +
  geom_point(aes(v1, v2)) +
  geom_line(aes(x, y), data = log_line(v1, v2)) +
  coord_trans(y = "log")

ggplot2 log scale of y axis causing curved lines

1 Answers