[R] plot scaling bug when using par(mfg)

Mark Cowley m.cowley at garvan.org.au
Thu Apr 30 07:42:39 CEST 2009

Dear list,
I've noticed a width/height scaling bug in multi-panel plots when you  
change the active plot using par mfg
The underlying par setting that maps the full plot width (or height)  
to the numerical axis labels does not get updated when you change  
plots with par(mfg=...). Thus plotting a point at c(10,10) in the  
first plot will only be in the correct place IF the most recently  
generated plot had the same x and y limits. This affects all plots  
except the most recently created one.

The error is reproduced whether you set the plot up using:  
par(mfcol=c(1,3)) -or- par(mfrow=c(1,3)) -or- layout(matrix(c(1,3),1,2))
I've tested this with 2x2, 1x2, 1x3 plots and all reproduce the error
Tested on R 2.8.0 on OSX and R 2.7.1 on WinXP and both reproduce the  

See ### comments for the errors below.


# draw 3 simple plots
plot(1:100, 1:100)
plot(1:200, 1:200)
plot(1:300, 1:300)

# add points to first plot which should span the whole plot
points(1:100, 1:100, pch="x", col="red")
# you should only see one dotted line at 100
abline(v=c(100,200,300), h=c(100,200,300), lty=c("dotted", "dashed",  
### instead you see 3 lines at 33, 66, 100

# add points to second plot which should span the whole plot
points(1:200, 1:200, pch="x", col="red")
# you should see a dotted and dashed line at 100 and 200
abline(v=c(100,200,300), h=c(100,200,300), lty=c("dotted", "dashed",  
### instead you see 3 lines at 66, 133, 200

# add points to third plot which should span the whole plot
points(1:300, 1:300, pch="x", col="red")
abline(v=c(100,200,300), h=c(100,200,300), lty=c("dotted", "dashed",  
### there are indeed 3 lines at 100, 200, 300

R version 2.8.0 (2008-10-20)


attached base packages:
[1] graphics  grDevices datasets  utils     stats     methods   base

other attached packages:
  [1] pwbc_0.1        mjcdev_0.1      tigrmev_0.1     slfa_0.1         
  [6] qtlreaper_0.1   pajek_0.1       mjcstats_0.1    mjcspot_0.1      
[11] mjcaffy_0.1     haselst_0.1     geomi_0.1       geo_0.1          
[16] cor_0.1         bootstrap_0.1   blat_0.1        bitops_1.0-4     
[21] gdata_2.4.2     gtools_2.5.0

Mark Cowley, PhD

Peter Wills Bioinformatics Centre
Garvan Institute of Medical Research, Sydney, Australia

More information about the R-help mailing list