Summarise x values at every unique y

[Simon]

I would like to display a line going through the means of x values for
every unique y value:
stat_summary gives me the opposite (as documented...) :
qplot(displ, reorder(manufacturer, displ), data = mpg)+stat_summary(aes
(group=1), fun.y=mean, geom="line", col=2)

ie, the equivalent of this:
qplot(reorder(manufacturer, displ), displ, data = mpg)+stat_summary(aes
(group=1), fun.y=mean, geom="line", col=2)+coord_flip()

except that for my real data I use a facet_grid (facet_grid(td~.,
scale="free", space="free", as.table=FALSE)) and the coord_flip seems
to mess it up...

How can I achieve this?

TIA

Simon.

[Simon]

I forgot to cc to the list an example with real data that Juliet
requested... so here it is:

Using the mpg dataset available in the ggplot2 package:
data(mpg)
# I would like to add to the following plot, lines, in each panel,
that go through the means of "displ" for each car
qplot(displ, reorder(model, displ), data = mpg)+facet_grid
(manufacturer~., space="free", scales="free_y")

# As is expect from the documentation, the following line computes
"column" means
# instead of the "row" means I'd like
qplot(displ, reorder(model, displ), data = mpg)+stat_summary(aes
(group=1), fun.y=mean, geom="line", col=2)+facet_grid(manufacturer~.,
space="free", scales="free_y")

# And if I try to rotate the data, this somehow seems to mess the
scales of the panels:
qplot(reorder(model, displ), displ, data = mpg)+stat_summary(aes
(group=1), fun.y=mean, geom="line", col=2)+facet_grid(manufacturer~.,
space="free", scales="free_x")+coord_flip()

So I am stumped...

Any Idea?

Thanks,

Simon.

[hadley wickham]

library(ggplot2)


Whenever the built in ggplot2 summaries can't do what you want, it's
usually best to start doing it yourself. You can compute the per
model displacement means as follows:

means <- ddply(mpg, c("manufacturer", "model"), summarise,
displ = mean(displ))

You can then add it to your plot as follows:

qplot(displ, reorder(model, displ), data = mpg) +

geom_point(data = means, colour = "red", size = 2) +
facet_grid(manufacturer~., space="free", scales="free_y")

Regards,

Hadley

--
http://had.co.nz/

[Simon de Bernard]

Hi Hadley,

thanks for your answer, however my problem is not in computing and displaying the means, but with drawing lines through the means (see my examples)

Any idea?

Thanks,

Simon.

[hadley wickham]

qplot(displ, reorder(model, displ), data = mpg) +

geom_vline(aes(xintercept = displ), data = means, colour = "red") +

facet_grid(manufacturer~., space="free", scales="free_y")

?

Hadley

--
http://had.co.nz/

[Simon de Bernard]

Sorry I wasn't clear... when I said through the means I meant connecting the means...

Like this:

qplot(reorder(model, displ), displ, data = mpg)+stat_summary(aes 
(group=1), fun.y=mean, geom="line", col=2)+facet_grid(manufacturer~., 
space="free", scales="free_x")+coord_flip() 

except that facet_grid seems to disregard the 'scales="free_x"' parameter...

Simon

[hadley wickham]

Oh, I finally get it - sorry for being so dense. Basically what you
want is 90 degree rotation of geom_line - you can get that with
coord_flip, but coordinate systems do not currently interact well with
free scales. An alternative is to use geom_path, which connects
points in the in order which they appear in the data:

library(ggplot2)

means <- ddply(mpg, c("manufacturer", "model"), summarise,
displ = mean(displ))

means <- means[with(means, order(manufacturer, displ)), ]

qplot(displ, reorder(model, displ), data = mpg) +

geom_path(aes(group = 1), data = means, colour = "red") +

facet_grid(manufacturer~., space="free", scales="free_y")

Regards,

Hadley

On Sat, Sep 5, 2009 at 3:38 PM, Simon de

--
http://had.co.nz/

[Simon de Bernard]

That's it :-)

Thanks

Simon.