Nested facets
Michael Kubovy
University of Virginia
Department of Psychology
for mail add: for FedEx or UPS add:
room phone
Jean-Olivier Irisson
It seems to me that the cask letter is just a label, rather than a coordinate, and that you just want the casks ordered by increasing strength in each batch. In which case you may want to mask the y coordinate and label points with a colour or a shape (or forget about cask completely).
data(Pastes)
# reorder batches
Pastes$batch <- reorder(Pastes$batch, -Pastes$strength)
# compute the order of the casks within each batch
library(plyr)
Pastes <- ddply(Pastes, ~batch, function(x) {
x$caskScore <- as.numeric(reorder(x$cask, x$strength))
return(x)
})
# label the casks with a colour/shape
ggplot( Pastes, aes( y = factor(caskScore), x = strength , shape=cask) ) + geom_point() + facet_grid(batch~.)
# NB: ideally you would hide only the y axis but that's not possible with the current theming system
# NB: using colour here prevents you from using geom_smooth because the smooth would be done by colour, but geom_smooth is probably not what you want here anyway
Another possibility is to create a fake y coordinate and relabel if afterwards
data(Pastes)
# reorder batches
Pastes$batch <- reorder(Pastes$batch, -Pastes$strength)
# compute the order of the casks within each batch
library(plyr)
Pastes <- ddply(Pastes, ~batch, function(x) {
x$caskScore <- as.numeric(reorder(x$cask, x$strength))
return(x)
})
# compute an overall y scale, ranking by batch (tens) and then cask within each batch (unities)
Pastes$batchScore <- as.numeric(Pastes$batch)
Pastes$score <- Pastes$batchScore*10 + Pastes$caskScore
ggplot( Pastes, aes( y = score, x = strength) ) + geom_point() +
# the facets do not have the same scale so forget about it
facet_grid(batch~., scales="free_y") +
# relabel the y axis breaks
scale_y_continuous("Cask within batch", breaks=Pastes$score, labels=Pastes$sample, expand=c(0.05, 0.2))
# you also jitter points along the y axis but I don't see why actually
# you could do it this way
ggplot( Pastes, aes( y = score, x = strength) ) + geom_point(position = position_jitter( height = 0.05 )) +
# the facets do not have the same scale so forget about it
facet_grid(batch~., scales="free_y") +
# relabel the y axis breaks
scale_y_continuous("Cask within batch", breaks=Pastes$score, labels=Pastes$sample, expand=c(0.05, 0.2))
# except that removes some y breaks (<- no idea why...)
# now you want to add the lines, which are just connecting the mean for each sample if I understand correctly
# compute the mean per sample (but keep all other info - batch, cask etc. - for the plot)
PastesAvg <- ddply(Pastes, ~batch+cask+sample+score, function(x) mean(x$strength))
# plot everything
ggplot( Pastes, aes( y = score, x = strength) ) + geom_point() +
geom_line(aes(x=V1, y=score), data=PastesAvg) +
facet_grid(batch~., scales="free_y") +
scale_y_continuous("Cask within batch", breaks=Pastes$score, labels=Pastes$sample, expand=c(0.05, 0.2))
I think this is exactly the plot you want, except from minor aesthetic aspects (in a future version of ggplot, you'll be able to put the y axis labels on the right, next to the facet labels).
Jean-Olivier Irisson
---
Observatoire Océanologique
Station Zoologique, B.P. 28, Chemin du Lazaret
06230 Villefranche-sur-Mer
Tel: +33 04 93 76 38 04
Mob: +33 06 21 05 19 90
http://jo.irisson.com/
Dennis Murphy
A little bit tricky, but I think this is what you were after:
library('plyr')
Pastes <- ddply(Pastes, 'ss', mutate, avgstr = mean(strength),
bb2 = factor(bb, levels = rev(levels(bb))))
ggplot(Pastes, aes(y = reorder(sample, strength, mean))) +
theme_bw() +
geom_point(aes(x = strength), color = 'dodgerblue') +
geom_line(aes(x = avgstr, group = 1), color = 'dodgerblue') +
facet_grid(bb2 ~ ., scales = 'free_y') +
labs(x = 'Paste strength', y = 'Sample within batch') +
opts(strip.background = theme_rect(fill = 'peachpuff'))
It's less of a headache if you compute the sample within batch means
outside of ggplot2; I used the mutate() function from plyr to generate
the means as well as reversing the levels of bb, since as.table didn't
work for me in facet_grid() - I got the same graph with as.table =
TRUE and as.table = FALSE, where the strengths were low at the top and
high at the bottom.
Notice that the x inputs in geom_point() and geom_line() are
different: the strength points are plotted, but the mean strengths are
connected with lines. In this plot, strength is the x and ordered
factor level is the y, so the group = 1 aesthetic is necessary to
connect the points across factor levels. This is how you get the
equivalent of panel.average() in ggplot2. To get separate y labels per
facet, use scales = 'free_y'.
The only obvious difference is that the strip labels are on the right
rather than the left; if there's a way to do it in ggplot2, someone
else will have to chime in because I don't know how to do it at
present, short of hacking into the facet_grid() code.
HTH,
Dennis
> --
> You received this message because you are subscribed to the ggplot2 mailing
> list.
> Please provide a reproducible example:
> https://github.com/hadley/devtools/wiki/Reproducibility
>
> To post: email ggp...@googlegroups.com
> To unsubscribe: email ggplot2+u...@googlegroups.com
> More options: http://groups.google.com/group/ggplot2
Hadley Wickham
> outside of ggplot2; I used the mutate() function from plyr to generate
> the means as well as reversing the levels of bb, since as.table didn't
> work for me in facet_grid() - I got the same graph with as.table =
> TRUE and as.table = FALSE, where the strengths were low at the top and
> high at the bottom.
That's https://github.com/hadley/ggplot2/issues/497 and should be
fixed in the next version.
Hadley
--
Assistant Professor / Dobelman Family Junior Chair
Department of Statistics / Rice University
http://had.co.nz/