R help - Sep 2009 - ggplot2: deterministic position_jitter & geom_line with position_jitter

Dear guRus,

I am starting to work with the ggplot2 package and have two very dumb 
questions:

1) deterministic position_jitter - the jittering is stochastic; is there 
any way to get a deterministic jittering? For instance:

example.data <- 
data.frame(group=c("foo","bar","foo","bar","foo","bar"),x=c(1,1,2,2,3,3),y=c(1,1,0,2,1,1))
set.seed(2009)
qplot(x,y,data=example.data,shape=group,position=position_jitter(w=0.1,h=0))

For x=1, the foo point is to the left of the bar point, and for x=3 the 
other way around. I would like to have all foo points at 
seq(1,3)-epsilon and all bar points at seq(1,3)+epsilon. Do I need to 
manually modify example.data$x groupwise for this?

2) geom_line with position_jitter - when I call multiple geoms with 
position_jitter, each geom gets its own jittering. For example 
(continuing with example.data above):

set.seed(2009)
qplot(x,y,data=example.data,geom=c("point","line"),shape=group,position=position_jitter(w=0.1,h=0))

The lines do not connect the points - is there any way to have the 
geom_line connect all the foo points on the one hand and all the bar 
points on the other hand?

--------------------------------------------------------------

What I've done: searched through HW's book, googled, searched RSeek. For
point 2) above, I tried using multiple layers and resetting the seed in 
between, to wit:

pp <- ggplot(example.data,aes(x,y,shape=group))
set.seed(2009)
pp <- pp+layer(geom="point",position=position_jitter(w=0.1,h=0))
set.seed(2009)
pp <- pp+layer(geom="line",position=position_jitter(w=0.1,h=0))
print(pp)

This doesn't do what I want, either...

Why I'm doing this: example.data actually contains group means across a 
covariate x, and they need to be plotted as "dots plus error bars" 
(psychologists' convention), so "use boxplots instead" is a
perfectly
correct reply to my questions which unfortunately does not help me.

Thanks for your time!
Stephan

----------------------------------------------------------------------

sessionInfo():

R version 2.9.2 (2009-08-24)
i386-pc-mingw32

locale:
LC_COLLATE=German_Germany.1252;LC_CTYPE=German_Germany.1252;LC_MONETARY=German_Germany.1252;LC_NUMERIC=C;LC_TIME=German_Germany.1252

attached base packages:
  [1] grid      grDevices datasets  splines   graphics  stats     tcltk 
     utils     methods   base

other attached packages:
  [1] ggplot2_0.8.3   reshape_0.8.3   plyr_0.1.9      proto_0.3-8 
svSocket_0.9-43 svMisc_0.9-48   TinnR_1.0.3     R2HTML_1.59-1 
Hmisc_3.7-0
[10] survival_2.35-4

loaded via a namespace (and not attached):
[1] cluster_1.12.0  lattice_0.17-25 tools_2.9.2

On Sat, Sep 12, 2009 at 1:34 PM, Stephan Kolassa <Stephan.Kolassa at
gmx.de> wrote:
> Dear guRus,
>
> I am starting to work with the ggplot2 package and have two very dumb
> questions:
>
> 1) deterministic position_jitter - the jittering is stochastic; is there
any
> way to get a deterministic jittering? For instance:
>
> example.data <-
>
data.frame(group=c("foo","bar","foo","bar","foo","bar"),x=c(1,1,2,2,3,3),y=c(1,1,0,2,1,1))
> set.seed(2009)
>
qplot(x,y,data=example.data,shape=group,position=position_jitter(w=0.1,h=0))
>
> For x=1, the foo point is to the left of the bar point, and for x=3 the
> other way around. I would like to have all foo points at seq(1,3)-epsilon
> and all bar points at seq(1,3)+epsilon. Do I need to manually modify
> example.data$x groupwise for this?
Yes.  The plyr package generally makes this sort of manipulation pain free.
> 2) geom_line with position_jitter - when I call multiple geoms with
> position_jitter, each geom gets its own jittering. For example (continuing
> with example.data above):
>
> set.seed(2009)
>
qplot(x,y,data=example.data,geom=c("point","line"),shape=group,position=position_jitter(w=0.1,h=0))
>
> The lines do not connect the points - is there any way to have the
geom_line
> connect all the foo points on the one hand and all the bar points on the
> other hand?
Again, you'll have to adding the jittering yourself.
> --------------------------------------------------------------
>
> What I've done: searched through HW's book, googled, searched
RSeek. For
> point 2) above, I tried using multiple layers and resetting the seed in
> between, to wit:
>
> pp <- ggplot(example.data,aes(x,y,shape=group))
> set.seed(2009)
> pp <-
pp+layer(geom="point",position=position_jitter(w=0.1,h=0))
> set.seed(2009)
> pp <-
pp+layer(geom="line",position=position_jitter(w=0.1,h=0))
> print(pp)
>
> This doesn't do what I want, either...
The jittering isn't evaluated until render time, so this won't help.
You'll notice if you draw the plot again, you'll get a different
rendering.

Hadley

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