How to order breaks with ggplot / geom_bar

3
votes

I have a data.frame with entries like:

   variable  importance order
1       foo  0.06977263     1
2       bar  0.05532474     2
3       baz  0.03589902     3
4     alpha  0.03552195     4
5      beta  0.03489081     5
       ...

When plotting the above, with the breaks = variable, I would like for the order to be preserved, rather than placed in alphabetical order.

I am rendering with: 

 
ggplot (data, aes(x=variable, weight=importance, fill=variable)) + 
    geom_bar() + 
    coord_flip() + opts(legend.position='none')

However, the ordering of the variable names is alphabetical, and not the order within the data frame. I had seen a post about using "order" in aes, but appears to have no effect.

I am looking to have a breaks ordering in-line with the "order" column.

There seems to be a similar question How to change the order of discrete x scale in ggplot, but frankly, did not understand the answer in this context.

4
rggplot2
FYI: It's easier for people to answer if you use dput and add a small sample of the data to the end of your post (or else use a globally available dataset from the datasets package in base R). – Shane
Is variable a character or a factor? – Joshua Ulrich
variable is a factor. I'm going to look at the answers below ... – Jonathan Shore
Shane, thanks for the tip. Was not aware of dput. – Jonathan Shore

4 Answers

6
votes

Try:

data$variable <- factor(data$variable, levels=levels(data$variable)[order(-data$order)])

From: ggplot2 sorting a plot Part II

4
votes

Even shorter and easier to understand:

data$Variable <- reorder(data$Variable, data$order)
2
votes

Another solution is to plot the order and then change the labels after the fact:

df <- data.frame(variable=letters[c(3,3,2,5,1)], importance=rnorm(5), order=1:5)
p <- qplot(x=order, weight=importance, fill=variable, data=df, geom="bar") + 
  scale_x_continuous("", breaks=1:5, labels=df$variable) + 
  coord_flip() + opts(legend.position='none')
1
votes

A shot in the dark, but maybe something like this:

data$variable <- factor(data$variable, levels=data$variable)