How to plot errorbars on this plot and change the overlay? - r

Question

Hi have this dataset :
tdat=structure(list(Condition = structure(c(1L, 3L, 2L, 1L, 3L, 2L,
1L, 3L, 2L, 1L, 3L, 2L, 1L, 3L, 2L, 1L, 3L, 2L, 1L, 3L, 2L, 1L,
3L, 2L, 1L, 3L, 2L), .Label = c("AS", "Dup", "MCH"), class = "factor"),
variable = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L,
3L, 3L, 3L), .Label = c("Bot", "Top", "All"), class = "factor"),
value = c(1.782726022, 1, 2.267946449, 1.095240234, 1, 1.103630141,
1.392545278, 1, 0.854984833, 4.5163067, 1, 4.649271897, 0.769428018,
1, 0.483117123, 0.363854608, 1, 0.195799358, 0.673186975,
1, 1.661568993, 1.174998373, 1, 1.095026419, 1.278455823,
1, 0.634152231)), .Names = c("Condition", "variable", "value"
), row.names = c(NA, -27L), class = "data.frame")
> head(tdat)
Condition variable value
1 AS Bot 1.782726
2 MCH Bot 1.000000
3 Dup Bot 2.267946
4 AS Bot 1.095240
5 MCH Bot 1.000000
6 Dup Bot 1.103630
I can plot it like that using this code :
ggplot(tdat, aes(x=interaction(Condition,variable,drop=TRUE,sep='-'), y=value,
fill=Condition)) +
geom_point() +
scale_color_discrete(name='interaction levels')+
stat_summary(fun.y='mean', geom='bar',
aes(label=signif(..y..,4),x=as.integer(interaction(Condition,variable))))
I have 2 questions :
How to change the overlay so the black points are not hidden by the
bar chart (3points should be visible per column)
How to add vertical errorbar on top of the bars using the standard
deviation from the black points ?

Answer 1

I'm not much in favor of mixing error bars with a bar plot.
In ggplot2 geoms are drawn in the order you add them to the plot. So, in order to have the points not hidden, add them after the bars.
ggplot(tdat, aes(x=interaction(Condition,variable,drop=TRUE,sep='-'), y=value,
fill=Condition)) +
stat_summary(fun.data="mean_sdl", mult=1, geom="errorbar") +
stat_summary(fun.y='mean', geom='bar') +
geom_point(show_guide=FALSE) +
scale_fill_discrete(name='interaction levels')

Answer 2

Like this:
tdat$x <- with(tdat,interaction(Condition,variable,drop=TRUE,sep='-'))
tdat_err <- ddply(tdat,.(x),
summarise,ymin = mean(value) - sd(value),
ymax = mean(value) + sd(value))
ggplot(tdat, aes(x=x, y=value)) +
stat_summary(fun.y='mean', geom='bar',
aes(label=signif(..y..,4),fill=Condition)) +
geom_point() +
geom_errorbar(data = tdat_err,aes(x = x,ymin = ymin,ymax = ymax,y = NULL),width = 0.5) +
labs(fill = 'Interaction Levels')
I've cleaned up your code somewhat. You will run into fewer problems if you move any extraneous computations outside of your ggplot() call. Better to create the new x variable first. Everything is more readable that way too.
The overlaying issue just requires re-ordering the layers.
Note that you were using scale_colour_* when you had mapped fill not colour (this is a very common error).
The only other "trick" was the un-mapping of y. Normally, when things get tricky I omit aes from the top level ggplot call entirely to make sure that each layer gets only the aesthetics that it needs.
The error bars again I tend to create the data frame outside of ggplot first. I find that cleaner and easier to read.