Multiple rugs

[Stavros Macrakis]

How do I go about creating multiple (non-overlapping) rugs?  + geom_rug(...) + geom_rug(...) puts them both in the same space.  I understand that rug lines can have multiple colors, but sometimes the density of the data is such that multiple rugplots would be helpful....

            -s

[baptiste auguie]

Hi,

It seems hard-coded, but you could slightly modify the geom and create your own like this,


GeomRug <- proto(Geom, {
  draw <- function(., data, scales, coordinates, ...) {
    rugs <- list()
    data <- coordinates$transform(data, scales)
    if (!is.null(data$x)) {
      rugs$x <- with(data, segmentsGrob(
        x0 = unit(x, "native"), x1 = unit(x, "native"),
        y0 = unit(ystart, "npc"), y1 = unit(ystart, "npc") + unit(ylen, "npc"),  # FIRST MODIF
        gp = gpar(col = alpha(colour, alpha), lty = linetype, lwd = size * .pt)
      ))
    }

    if (!is.null(data$y)) {
      rugs$y <- with(data, segmentsGrob(
        y0 = unit(y, "native"), y1 = unit(y, "native"),
        x0 = unit(xstart, "npc"), x1 = unit(xstart, "npc") + unit(xlen, "npc"),  # SECOND MODIF
        gp = gpar(col = alpha(colour, alpha), lty = linetype, lwd = size * .pt)
      ))
    }

    gTree(children = do.call("gList", rugs))
  }

  objname <- "rug"

  desc <- "Free rug plots"

  default_stat <- function(.) StatIdentity
 
  default_aes <- function(.) # THIRD MODIF
        aes(colour="black", size=0.5, linetype=1, alpha = 1, ystart=0, ylen=0.03, xstart=0, xlen=0.03) 
  guide_geom <- function(.) "path"

  examples <- function(.) {
    p <- ggplot(mtcars, aes(x=wt, y=mpg))
    p + geom_point()
    p + geom_point() + geom_rug()
    p + geom_point() + geom_rug(position='jitter')
  }


})


p <- ggplot(mtcars, aes(x=wt, y=mpg))

p + geom_point() + geom_rug()
p + geom_point() + geom_rug(xstart=0.5)

I'm not sure how to rename it as a new geom though, and it would need some thoughts about stacking several rugs on top of each other automatically (whether that's a good idea or not I dunno).

HTH,

baptiste

> Physics
University of Exeter
Stocker Road,
Exeter, Devon,
EX4 4QL, UK

Phone: +44 1392 264187

http://newton.ex.ac.uk/research/emag
______________________________

---

From: Stavros Macrakis <macr...@gmail.com>
To: ggplot2 <ggp...@googlegroups.com>
Sent: Thursday, 7 May, 2009 22:28:27
Subject: Multiple rugs

How do I go about creating multiple (non-overlapping) rugs?  + geom_rug(...) + geom_rug(...) puts them both in the same space.  I understand that rug lines can have multiple colors, but sometimes the density of the data is such that multiple rugplots would be helpful....

            -s

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the ggplot2 mailing list.
To post to this group, send email to ggp...@googlegroups.com
To unsubscribe from this group, send email to
ggplot2+u...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/ggplot2
-~----------~----~----~----~------~----~------~--~---

[hadley wickham]

I've never found much use for rug plots. But if Baptiste's suggestion
doesn't work, you can always draw your own with geom_segment.

Hadley

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

[Stavros Macrakis]

I'm not particularly fond of rug plots, either.  But until someone
(maybe me if I can figure out how...) adds density plots to the axes,
they're a quick and dirty way of seeing density along each axis.  When
they get too dense, I sometimes will sample from the original
distribution to generate a more legible rugplot....

               -s

[Peter Flom]

Stavros Macrakis <macr...@gmail.com> wrote

>I'm not particularly fond of rug plots, either.  But until someone
>(maybe me if I can figure out how...) adds density plots to the axes,
>they're a quick and dirty way of seeing density along each axis.  When
>they get too dense, I sometimes will sample from the original
>distribution to generate a more legible rugplot....
>

I've lost the origin of this thread, but for adding box plots to each axis
of a scatterplot there is the scatterplot function in the car package, which does other
nice things as well.

Peter

Peter L. Flom, PhD
Statistical Consultant
www DOT peterflomconsulting DOT com

[Mike Lawrence]

Here's my attempt at adding density plots to each axis. Unfortunately,
the density plots don't properly line up with the scatterplot and
there's a lot of white space between the scatter and density plots.

d = data.frame(V1 = rnorm(100),V2 = rnorm(100))
p1 = ggplot()+
layer(
data = d
, mapping = aes(
x = V1
, y = V2
)
, geom = 'point'
)+
scale_x_continuous(
limits = c(min(d$V1),max(d$V1))
)

p2 = ggplot()+
layer(
data = d
, mapping = aes(
x = V1
, y = ..density..
)
, geom = 'line'
, stat = 'density'
)+
scale_x_continuous(
limits = c(min(d$V1),max(d$V1))
)+
opts(
panel.grid.minor = theme_blank()
,panel.grid.major = theme_blank()
,axis.ticks = theme_blank()
,axis.text.y = theme_blank()
,axis.text.x = theme_blank()
,axis.title.y = theme_blank()
,axis.title.x = theme_blank()
)

p3 = ggplot()+
layer(
data = d
, mapping = aes(
x = V2
, y = ..density..
)
, geom = 'line'
, stat = 'density'
)+
scale_x_continuous(
limits = c(min(d$V2),max(d$V2))
)+
opts(
panel.grid.minor = theme_blank()
,panel.grid.major = theme_blank()
,axis.ticks = theme_blank()
,axis.text.y = theme_blank()
,axis.text.x = theme_blank()
,axis.title.y = theme_blank()
,axis.title.x = theme_blank()
)+
coord_flip()

size = .7
print(p1,vp = viewport(width = size , height = size , x = size*.5 , y
= size*.5 ))
print(p2,vp = viewport(width = size , height = 1-size , x = size*.5 ,
y = 1-(1-size)*.5 ))
print(p3,vp = viewport(width = 1-size , height = size , x =
1-(1-size)*.5 , y = size*.5 ))

--
Mike Lawrence
Graduate Student
Department of Psychology
Dalhousie University

Looking to arrange a meeting? Check my public calendar:
http://tr.im/mikes_public_calendar

~ Certainty is folly... I think. ~