| • | basic R | • | data input/output | • | R packages | • | plotting/graphics | • | R scripts | • | using seeds | • | more info | • |
| Basic R | [top] |
If you are in Unix/Linux, R uses directory-specific workspaces (unlike S-Plus, for example, which will load the same global workspace, no matter from where you start it).
Here are some basic R commands to get you started.
| R I/O | [top] |
| dump("variable",file="variable.R") | dumps a given variable out into a file (to be read in by source or scan) |
| load(<R file>) | loads an R object saved with save |
| save("variable", file="variable.R") | saves an R object (to be loaded with load). This is supposed to be "safer" than eg dump. |
| scan("file.R") | reads an external input file into a variable (eg double data from Matlab) |
| sink() | writes R console text (to eg a file) |
| source("file.R") | load functions/R code into the workspace |
| R Packages | [top] |
I have co-written a semi-successful help page for creating and installing R packages. Many thanks to Christian Stamm who has summarized and translated our guide for the German speaker. If you are Italian, then there is a guide by Fabio Frascati on CRAN.
| Plotting/graphics | [top] |
Devices
| dev.copy(device=X11) | makes an exact (X11) copy of the current device |
| dev.copy2eps(file="copy.eps") | copies the current device to an external file (copy.eps) |
| dev.cur() | contains value of the current device |
| dev.list() | lists all open devices |
| dev.off(a) | closes device number a |
| dev.set(a) | changes current (working) device to device number a |
| graphics.off() | closes all open (graphics) devices |
| get(getOption("device"))() | opens a graphics device (not platform-specific) |
| motif() | open a graphics device |
| X11() | open a graphics device |
| postscript("file.ps",horizontal=FALSE,width=8, height=6.25) | creates an "external" postscript file of specified width and height (cm) [listed dimensions are "good" for journal figures]. Device should be closed with dev.off. The commands png and pdf *should* work as well for graphics creation. |
Plotting
| legend(x,y,legend,fill=v) | adds legend to plot with top left corner at (x,y) according to axes with key vector legend and colour description v |
| par(cex=0.5) | changes size of graphical text/symbols (globally) |
| par(mfg=c(a,b,c,d)) | sets current device to position (a,b) in a (c,d) multiple figure environment |
| par(mfcol=c(a,b)) | opens an empty multiple figure environment (axb,filled column by column) |
| par(mfrow=c(a,b)) | opens an empty multiple figure environment (axb,filled row by row) |
| par(xpd=NA) | enables writing in plot margins |
| plot(...,cex=0.5) | changes size of graphical text/symbols (for current plot) |
For more detailed information about graphics and plotting techniques in R, there is a very good book, cunningly titled R Graphics by Paul Murrell.
| R scripts | [top] |
A script is a useful way in Linux of running e.g. R simulations in the background. The form of the executable is like any other shell script. Here are some script argument calls for R:
| R | --save* | saves workspace at end of a script |
| --no-save* | doesn't save workspace at end of script | |
| --restore | restores previously stored workspace | |
| --no-restore | doesn't restore previously stored workspace | |
| --q | no introductory information (version etc) |
* to be able to dump data out from a workspace into a directory from an R script (e.g. using dump), you need to have either --save or --no-save (see the example below).
The input file to R (named inscone in the shell script example) in a script is just a list of R commands. Here is an example:
| date() | ||
| library(roots) | ||
| x<-abs(runif(10)) | ||
| ans<-fifthroot(x) | ||
| dump("ans",file="ans.txt") | ||
| date() | ||
| q("yes") | &larr | If you have a q("yes") or q("no") included at the end of your R script, this will override the specification of --save or --no-save |
| Using seeds | [top] |
R is strange, and sometimes produces the same "random" numbers if asked, for example when using rnorm or runif. To avoid pesky duplicate sets of random numbers, you could do something similar to the following (for example, for use in R scripts and simulations):
| s<-system("date '+%s'",TRUE) | &larr | sets the seed according to a numeric interpretation of the current date/time (in seconds). Although this is strictly a Linux command, it should also work in R for windows as well. The TRUE argument makes s usable by R. |
| s<-as.integer(s) | ||
| set.seed(s) |
| More info... | [top] |
More information about R can be found on CRAN and the R help mailing lists.