Question

I’d like to put my own labels on a log-transformed y-axis of a boxplot:

with(TX, boxplot(b~tx, ylab="Biomass, g/m2, ln", yaxt="n", las=2, log="y", cex.axis=0.7))

The data:

pretty(log(TX$b))

[1] -6 -4 -2  0  2  4  6

length(levels(TX$tx))

[1] 22

I tried:

axis(2, at=pretty(log(TX$b)), labels=pretty(log(TX$b))) and 

axis(2, at=c(-6, -3, -1, 0, 1, 3, 6), labels=c(-6, -3, -1, 0, 1, 3, 6)), 

but in both cases only the positive values (and corresponding ticks!) are displayed. What’s the reason and how do I get this fixed?

Answer 1

You have transformed the y variable in the call to plot using log = 'y'.

This means you pass the at component as values on the untransformed scale.

axis(2, at= exp(c(-6, -3, -1, 0, 1, 3, 6)), labels=c(-6, -3, -1, 0, 1, 3, 6))

A reproducible example

set.seed(1)
TX <- data.frame(tx = gl(2,3), y= rlnorm(600, mean = -1, sd = 3))
boxplot( y~tx, data = TX, log = 'y',yaxt="n", las=2,  cex.axis=0.7)
axis(2, at= exp(c(-6, -3, -1, 0, 1, 3, 6)), labels=c(-6, -3, -1, 0, 1, 3, 6))