Summary statistic for each facet

[Felix Schönbrodt]

Hi there,

in Lattice I used to dig into a facetted panel function, calculate the
correlation of each x-y scatterplot and print the correlation as an
annotation into each facet.

Is something similar possible in ggplot2? I know that I can compute
each facet's correlation outside of ggplot2 and do an annotation with
geom_text; however it would be more convenient if the calculations
were done automatically accordingly to the facets specified.

Thanks,
Felix

[hadley wickham]

> Is something similar possible in ggplot2? I know that I can compute
> each facet's correlation outside of ggplot2 and do an annotation with
> geom_text; however it would be more convenient if the calculations
> were done automatically accordingly to the facets specified.

It would be convenient, but to work generally it would require an
infrastructure within ggplot2 for performing any kind of data
manipulation that you can imagine. I don't think that this is a good
idea in general, so I recommend performing your manipulations outside
of ggplot2. The plyr package is a natural fit for this sort of
activity:

cors <- ddply(mydf, c("group", "trt"), summarise, cor = cor(x, y))

Hadley

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

[Felix Schönbrodt]

Hi Hadley,

thanks for your reply - that seems convenient enough! ;-)
I haven't delved into the plyr-package so far, maybe I should do now ...

Best,
Felix