Grouping a "other" group with R and ggplot2

Question

I'm trying to figure out, how to best deal with an "other" group or how to create one. I'm using R and ggplot2.

Data

As an example: I'm analyzing the german Bundestagswahl (general election, every four years). There are at least 38 parties, that voters could vote for in 2009 (see bundeswahlleiter.de for numbers). Let's say, that I have an mysql database bundestagswahl with a column party and a column votes. Now I want to do some plots using ggplot.

Normally I would write a sql query with votes per party. Next I would tell ggplot2 to draw a geom_bar. This would draw me at least 38 bars.

Now my question

How can I create an "other" group, e.g. all parties, which have less then 5% of all votes to not clutter up the graph? This way I would have only 5 bars (=less clutter).

Of course I could write my sql-query to ignore parties, that have not at least 5%, but this way my counts wouldn't sum up to 100%. Over 6 % of all votes went to parties, which didn't made it into the parlament.

How would you do it? Write another sql query? Or using reshape? Or some other way?

Update

To make it a bit clearer, what I want to do, here is a fake dataset of six parties (names A, B, C, D, E and F) and there number of votes:

> party <- c("A", "B", "C", "D", "E", "F")
> votes <- c(1000, 2000, 5, 1500, 15, 20)
> df <- data.frame(party, votes)
> df
  party votes
1     A  1000
2     B  2000
3     C     5
4     D  1500
5     E    15
6     F    20

Now I want to do two things:

1. relabel all parties, which have less then 100 votes to "other".
2. add up those "other" parties to one entry.

Wanted outcome:

  party votes
1     A  1000
2     B  2000
3 other    40
4     D  1500

How do I do this? Is maybe combine_factor of the packages reshape an option?

Answer 1

Here's how I would probably approach the problem:

df <- within(df, other.group <- ifelse(votes > 100, as.character(party), "Other"))
df <- within(df, relevel(factor(other.group), "Other"))

ggplot(df, aes(other.group, votes)) + geom_bar(stat = "identity")

Answer 2

This will first assign "other" to any group with less then 5% of the votes and then give you the aggregated totals. The non-"other" groups are just their original values, while the "other" group gets summed.

df$party2 <- ifelse(df$votes/sum(df$votes) < 0.05, 
                    "other", 
                    as.character(df$party))
df.aggr <- aggregate(votes ~ party2, data = df, FUN = sum)

Which gives:

  party2 votes
1      A  1000
2      B  2000
3      D  1500
4  other    40