Question

I have data in a csv file in the following format

,BC1,BC10,BC11
1,2432,420,18
2,276,405,56
3,119,189,110
4,90,163,140
5,206,280,200
6,1389,1080,1075
7,3983,3258,4878
8,7123,15828,28111
9,8608,48721,52576
10,9639,44725,55951
11,8323,45695,32166
12,2496,18254,26600
13,1524,8591,18583
14,7861,1857,1680
15,10269,5165,4618
16,13560,64636,63262

I acquire the data in the following way

data <- read.csv(file="file.csv",sep=",",header=TRUE)
data <- data[,2:ncol(dat)]

Then transform it

datam <- melt(cbind(data,ind=rownames(data)),is.var = c('bind'))

Then I create a stacked percent plot

ggplot(datam,aes(x = variable, y = value,fill = ind)) + 
geom_bar(position = "fill") + 
scale_y_continuous(labels = percent_format())

To which I get:

I am trying to order the values in the legend that should go (1…16) instead of 1,10,…,16,2…,9. The color order is correct in the legend though. Thanks

Answer 1

It’s because ind (your fill variable) is being sorted as a string factor.

You can convert ind to numeric and then use it as the fill argument:

datam$ind <- as.numeric(datam$ind)
ggplot(datam, aes(x=variable, y=value, fill=factor(ind))) + ...

However this does give factor(ind) as the title for the legend. You could do:

datam$ind <- factor(as.numeric(datam$ind))
ggplot(datam, aes(x=variable, y=value, fill=ind)) + ...

Or, you could just use scale_fill_discrete('My Legend Name') and avoid having to convert datam, e.g.:

ggplot(datam, aes(x=variable, y=value, fill=factor(as.numeric(ind)))) + ... +
      scale_fill_discrete('legend name')

Edit

The ordering of the legend labels is determined by levels(datam$ind):

> levels(datam$ind)
[1] "1"  "10" "11" "12" "13" "14" "15" "16" "2"  "3"  "4"  "5"  "6"  "7"  "8" 
[16] "9" 

That’s why you originally got the string-wise order.
By doing factor(as.numeric(ind)), levels(datam$ind) got re-sorted in numeric order.

So to specify them going from 16 to 1, say, you can do:

lvls <- levels(datam$ind) # "1" "10" "11" ...
lvls <- sort(as.numeric(lvls)) # 1 2 3 4 ..
lvls <- rev(lvls)              # 16 15 14 13 ...
# re-level datam$ind according to lvls:
datam$ind <- factor(datam$ind, levels=lvls)

# now we can plot!
ggplot(datam, aes(x=variable, y=value, fill=ind)) + ...

Hmm, that did change the colours/stacking order though. I’m unsure how to have the labels go one way but the stacking go the other — ggplot always has them in sync (ie if labels are 1:16, then so is the stacking order). Re switching the colours around, I’m sure there’s some ggplot way to reverse the colour scale, probably via scale_colour_manual (as that’s just a cosmetic change) – I don’t know how to do that myself though.