Dirk Eddelbuettel <[EMAIL PROTECTED]> writes: > On 18 June 2007 at 00:33, Roger Leigh wrote: > | The errbar() function does not work correctly on a log axis, due > | to the code assuming the axis is linear. > | > | Here is a tiny testcase to demonstrate: > | > | library(Hmisc) > | x <- c(0.03,0.1,0.3,1,3,10,30) > | y <- c(10,6,2,0.8,0.3,0.06,0.021) > | sd <- c(5,2,1,0.3,0.1,0.2,0.08) > | par(mfrow=c(2,1)) > | errbar(x,y,y-sd,y+sd) > | errbar(x,y,y-sd,y+sd, log="x") > | > | It should be possible for the code to check if par xlog=T and then > | adjust the x±cap to cope with the log scale.
This is how I solved it (very crudely). This code snippit assumes
vectors x, y (data series), and sd (the error). colour is the line
colour.
# Add error bars "by hand", for SD on a log axis.
i <- 1
for (i in 1: length(x)) {
lower <- 0.90 * x[i]
upper <- 1.12 * x[i]
lines(c(x[i],x[i]),c(y[i]+sd[i], y[i]-sd[i]), col=colour, lty=1)
lines(c(lower,upper), c(y[i]+sd[i], y[i]+sd[i]), col=colour, lty=1)
lines(c(lower,upper), c(y[i]-sd[i], y[i]-sd[i]), col=colour, lty=1)
}
It replaces "cap" with two factors for giving a nice even line over
the top and bottom of the error bar. However, this would ideally be
replaced by a computed value to allow the user to specify the cap on
some sort of meaningful scale.
Regards,
Roger
--
.''`. Roger Leigh
: :' : Debian GNU/Linux http://people.debian.org/~rleigh/
`. `' Printing on GNU/Linux? http://gutenprint.sourceforge.net/
`- GPG Public Key: 0x25BFB848 Please GPG sign your mail.
pgpp6TFXx7aDH.pgp
Description: PGP signature
