[R] Boxplot across levels of a factor

[Prof Brian Ripley]

On Tue, 24 Aug 2004, Ivar Herfindal wrote:

> I think Paul wanted one plot for each box, not all boxes in one plot (sorry 
> if I misunderstand).
> 
> One way to solve this can be like this:
> 
> par(mfrow=c(1,3))
> with(data, by(1:nrow(data), Type, function(x) boxplot(Locus[x]~data[x]))

Actually, that does not come close to working.  Did you try it?
(Please read the documentation for by().)  The following does work

by(data, Type, function(x) boxplot(x$Locus, 
                                   main=paste("Type", unique(x$Type))))

but that would be a very unusual plot, and one in which the three plots
had different scalings so the boxplots could not be compared.

For something like that, lattice's bwplot is a better option.

bwplot(Locus ~ Type, data=data)  1 panel
bwplot(Locus ~ factor(rep(1, 182)) | Type, data=data) 3 panels

> 
> Hope this works for you.
> 
> Ivar
> 
> On Tue, 24 Aug 2004 07:27:57 +0100 (BST), Prof Brian Ripley 
> <ripley at stats.ox.ac.uk> wrote:
> 
> > On Tue, 24 Aug 2004, Paul Boutros wrote:
> >
> >> Hello,
> >>
> >> I have a data-frame in which one-column is a factor:
> >>
> >> > str(data);
> >> `data.frame':   194 obs. of  8 variables:
> >> $ Type         : Factor w/ 3 levels "Nuclear-Rec..",..: 1 2 2 2 2 2 2 2 
> >> 2 2
> >> ...
> >> $ Locus        : num  0.000571 0.004000 0.001429 0.004857 0.007429 ...
> >>
> >> And I'd like to make a boxplot of the data$Locus values, where each 
> >> level of
> >> the factor gets its own box-and-whiskers plot.  I'm weak in R, but I 
> >> thought
> >> there might be some shortcut to automating this instead of just creating 
> >> a
> >> new data-structure with all the separate values?
> >
> > There are two.  The simpler is
> >
> > 	boxplot(Locus ~ Type, data=data)
> >
> > and you can also use
> >
> > 	with(data, boxplot(split(Locus, Type)))
> >
> > (split() does automate the construction of a suitable data structure.)
> >
> 
> 
> 
> 

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595