Order of ordered factors are not preserved in the boxplots.
Vitalie S.
Order of the factors changes while displayed on x axis.
Is that a feature?
fact<-factor(sample(c(1:5),20,rep=T),ordered=T)
num<-rnorm(20)
qplot(fact,num,geom="boxplot")
in plot they appear in order 43152 but
> str(fact)
Ord.factor w/ 5 levels "1"<"2"<"3"<"4"<..: 4 2 3 1 3 5 5 4 1 3 ...
Sorry if I missed something,
Thanks,
Vitalie.
Vitalie S.
Vitalie S.
> Is that a feature?
Probably a bug.
That appears to be a general way the ggplot2 works. In documentation for scales_discrete:
levels(diamonds$cut)
[1] "Fair" "Good" "Very Good" "Premium" "Ideal"
qplot(cut, data=diamonds, stat="bin")
levels in the plot are inverted and "Good" and "Very Good" swapped. Apparently for no reason.
A great deal of work!!
Thanks for the package.
Vitalie.
Christian Lindig
On Mar 13, 9:56 pm, "Vitalie S." <vitosm...@rambler.ru> wrote:
> > Order of the factors changes while displayed on x axis.
> > Is that a feature?
>
> Probably a bug.
> That appears to be a general way the ggplot2 works. In documentation for scales_discrete:
> levels(diamonds$cut)
> [1] "Fair" "Good" "Very Good" "Premium" "Ideal"
> qplot(cut, data=diamonds, stat="bin")
>
> levels in the plot are inverted and "Good" and "Very Good" swapped. Apparently for no reason.
of factors in general? Are labels and breaks the way to go and what is
the interaction between order and levels? I see the problem coming up
here repeatedly but I don't see the big picture.
-- Christian
jiho
> Using this opportunity: could someone explain how to control the order
> of factors in general?
> Are labels and breaks the way to go and what is
> the interaction between order and levels? I see the problem coming up
> here repeatedly but I don't see the big picture.
The order of factors is controlled by the factor/ordered functions
using the levels argument. The order is the one in which levels are
specified.
ggplot does not respect that currently but there is a work around
until Hadley fixes the issue (which is likely to be before you even
read this message).
data = data.frame(x=c("c","a","b","c","c","a"), stringsAsFactors =F)
qplot(data$x,geom="histogram")
# order of first occurence in the vector: c, a, b
class(data$x)
# data is still a character string
data$x = factor(data$x, levels=c("a","b","c"))
data$x
qplot(data$x,geom="histogram")
# should plot as a, b, c = the order of the levels
# I think this is a (recent) bug in ggplot. The behaviour was correct
previously IIRC.
# work around: physically order the data in the data.frame along the
order of the factor
data$x
data$x = data$x[order(as.numeric(data$x))]
data$x
qplot(data$x,geom="histogram")
# this works with any given order
data$x = factor(data$x, levels=c("b","c","a"))
data$x = data$x[order(as.numeric(data$x))]
qplot(data$x,geom="histogram")
Hope that helps.
JiHO
---
http://jo.irisson.free.fr/
Christian Lindig
Thanks a lot! This was indeed helpful and made me study the
documentation of factor() and especially the levels and labels
arguments. I think the role of levels and labels here is missing in
the discussion of scales, labels, and breaks in the ggplot2
documentation and should be clarified with an example that uses both
levels and labels. Levels are also important to label facets, which I
had asked about here previously and which was likewise not obvious for
me from the documentation.
-- Christian
hadley wickham
> using the levels argument. The order is the one in which levels are
> specified.
> ggplot does not respect that currently but there is a work around
> until Hadley fixes the issue (which is likely to be before you even
> read this message).
Yup, fixed in the development version.
> # work around: physically order the data in the data.frame along the
> order of the factor
> data$x
> data$x = data$x[order(as.numeric(data$x))]
> data$x
> qplot(data$x,geom="histogram")
> # this works with any given order
> data$x = factor(data$x, levels=c("b","c","a"))
> data$x = data$x[order(as.numeric(data$x))]
> qplot(data$x,geom="histogram")
Alternatively, you can set the limits on the x axis:
+ scale_x_discrete(limits = c("b","c","a"))
Hadley