Question

I am following this example to create polygons to plot using ggplot I am able to follow that example and create separate convex hulls if my data is subsetted; however, when I try to apply ddply as I have a grouping variable I am unable to. Here’s the code from the example with a added grouping variable :

library(grDevices) # load grDevices package
df <- data.frame(X = c(-62,  -40,   9,  13,  26,  27,  27),
             Y = c( 7, -14,  10,   9,  -8, -16,  12), id = c(1, 1, 1, 2, 2, 3, 3))
con.hull.pos <- ddply(df, .(id), summarize, hullpos = chull(X, Y)) # get convex hull positions by each ID

Now, to get a complete polygon for each ID, we need to get all the rows by each ID as given in con.hull.pos but we also need to add the first row of each group.

df[ddply(con.hull.pos, .(id), function(x) x[1, ])$hullpos, ] # first row of position
df[con.hull.pos$hullpos  ,]  ## all rows of position
rbind(df[con.hull.pos$hullpos  ,] , df[ddply(con.hull.pos, .(id), function(x) x[1, ])$hullpos, ])

and here my code fails as the first row using ddply is not the same as first row from the convex hull for an ID. Thus, the polygon is not complete. Can anyone please help me in applying the given example by grouping by a variable.

When subsetting manually, this code works as it creates three separate polygons covering three id areas

id1_df <- subset(df, id==1)
id1_con.hull.pos <- chull(id1_df$X, id1_df$Y)   

id2_df <- subset(df, id==2)
id2_con.hull.pos <- chull(id2_df$X, id2_df$Y)

id3_df <- subset(df, id==3)
id3_con.hull.pos <- chull(id3_df$X, id3_df$Y)

id1_con.hull <- rbind(id1_df[id1_con.hull.pos,], id1_df [id1_con.hull.pos[1],])
id2_con.hull <- rbind(id2_df [id2_con.hull.pos ,], id2_df  [id2_con.hull.pos [1],])
id3_con.hull <- rbind(id3_df [id3_con.hull.pos,], id3_df [id3_con.hull.pos[1],])             

poly_borders <- rbind(id1_con.hull, id2_con.hull, id3_con.hull)     

plot(Y ~ X, data = df) # plot data
lines(poly_borders) # add lines for convex hull

Answer 1

Two hints:

1. Use the chull function in base R to compute the convex hull.
2. Use dlply to store the resulting chull in a list, rather than a data frame

Then your code turns to:

x <- dlply(df, .(id), function(piece)piece[chull(piece$X, piece$Y), -3])

plot(Y~X, df)
lapply(x, polygon)

Which produces this plot:

---

If you want to plot this in ggplot it’s even easier, but use ddply:

x <- ddply(df, .(id), function(piece)piece[chull(piece$X, piece$Y), ])
ggplot(x, aes(X, Y, group=id)) + geom_polygon(fill="cyan", colour="blue") + geom_line()