user45017 user45017 -4 years ago 176
R Question

R: Sorting factor by level frequency and plotting

I have a data.frame with several factors such as:

df<-data.frame(Var1=as.factor(sample(c("AB", "BC", "CD", "DE", "EF"), 1000, replace=TRUE)))


209 195 178 221 197

I want to plot the frequency of the levels of each factor in the data.frame as follows:

ggplot(df, aes(x=factor(1), fill=factor(Var1)))+
geom_bar(width=1, colour="black")+

However, the order of the levels is alphabetically and not by frequency. Using count from library(plyr) I can create a new data.frame that gives me the frequency of each level:

df_count <-count(df, "Var1")
Var1 freq
1 AB 209
2 BC 195
3 CD 178
4 DE 221
5 EF 197

Which I can then reorder using

df_count$Var1<-factor(df_count$Var1, levels=df_count$Var1[order(df_count$freq, decreasing=TRUE)])

which when plotted give me what I want, the sorted frequency of each level.

1.) Is this the most elegant solution? It gives me an extra data.frame for each factor/column in my original data.frame, and I feel there must be a simpler way.

2.) When plotting, how can I rename the legend lables and ensure they are allocated the right factor level? If I use

scale_fill_manual(labels=c("Name of AB", "Name of BC", "Name of CD", "Name of DE","Name of EF"))

the labels do not relate to the right level. Here the first entry in the legend will be "DE" as it is the level with the highest frequency but the label will say "Name of AB" as defined in scale_fill_manual. I could check the order of the labels manually each time but there must be an automatic way?

Answer Source

You want reorder(), I think. In general reorder(x,y,FUN) changes the order of levels of a factor x according to the results of applying function FUN to a second variable y. In this case, you can use FUN=length, and it doesn't really matter what you use for y.

Set up data (I chose to modify the probabilities to make the result clearer):

df <- data.frame(Var1=as.factor(sample(c("AB", "BC", "CD", "DE", "EF"),
                                   1000, replace=TRUE)))

Basic plot (wrong order):

print(g1 <- ggplot(df, aes(x=factor(1), fill=Var1))+
   geom_bar(width=1, colour="black")+

enter image description here

Now reorder:

df$Var1 <- reorder(df$Var1,df$Var1,FUN=length)
## [1] "DE" "AB" "EF" "CD" "BC"

Check that the order is right:

##  DE  AB  EF  CD  BC 
##  46 105 163 189 497 

Print the new plot (stick in the new data with %+% and flip the order of the legend with guide_legend(): you could also use function(x) -length(x) as FUN to change the order of the levels in the first place).

print(g1 %+% df + 

enter image description here

Recommended from our users: Dynamic Network Monitoring from WhatsUp Gold from IPSwitch. Free Download