Question

I just discovered the great plyr package and am taking it for a spin.

A question I have is the following: is there some way to access the grouping variables from within d_ply?

Say I have a dataframe df with columns x,y,z, and I would like to plot for each z the values x versus y. If I do the following:

plotxy = function(df, ...) {plot(df$x, df$y, ...)}
d_ply(df, .(z), plotxy(df, main=.(z)))

then the titles that show up on the plots are all “z”, and not the values of the z variable. Is there a way to access those values from within d_ply?

EDIT: As @Justin pointed out, the above formulation is wrong because I am passing the whole of df to plotxy. Hence the line

d_ply(df, .(z), plotxy(df, main=.(z)))

should be

d_ply(df, .(z), plotxy, main=.(z))

in order to make sense in terms of my original question (I guess that’s also what @joran was hinting at).

However, I realized something else. Even though df gets sliced along z by d_ply, the sub-dataframe that the function receives still has a z column — simply with always the same value. Hence the problem can apparently be solved as follows:

plotxy = function(df, ...) {plot(df$x, df$y, main=df$z[1])}
d_ply(df, .(z), plotxy)

Answer 1

By way of example, I’ll expand on Joran’s concern.

df <- data.frame(x=rnorm(100), y=rnorm(100), z=letters[1:10])

lets use your function and see what we get without plyr:

plotxy(df, main=.(z))

versus the maybe more expected(?):

plotxy(df, main=df$z)

However, in you code, you are splitting your data frame on z then sending the whole data.frame df to your function again. Instead you could make a wrapper function:

d_ply(df, .(z), function(ply.df) plotxy(ply.df, main=unique(ply.df$z)))

This way the plotxy function is only seeing the smaller split data.frame ply.df that you pass through the wrapper function.