Plot average line in a facet_wrap

I have the following data set:

structure(list(Geschaeft = c(0.0961028525512254, 0.0753516756309475, 
0, 0.0722803347280335, 0, 0.000877706260971328), Gaststaette = c(0.0981116914423463, 
0.0789718659495242, 0.0336538461538462, 0.0905857740585774, 0, 
0.00175541252194266), Bank = c(0.100843712334271, 0.0717832023169218, 
0.00480769230769231, 0.025, 0.00571428571428572, 0.00965476887068461
), Hausarzt = c(0.0633989554037766, 0.0589573851882499, 0.0288461538461538, 
0.0217573221757322, 0.00685714285714286, 0.0128730251609128), 
    Einr..F..Aeltere = c(0.0337484933708317, 0.0550268928423666, 
    0.00480769230769231, 0, 0.00114285714285714, 0.000292568753657109
    ), Park = c(0.0738449176376055, 0.0726623913942904, 0.0625, 
    0.0846234309623431, 0.00228571428571429, 0.112053832650673
    ), Sportstaette = c(0.0449979911611089, 0.0612846503930492, 
    0.00480769230769231, 0.0619246861924686, 0.00114285714285714, 
    0), OEPNV = c(0.10847730012053, 0.089056681836988, 0.264423076923077, 
    0.135669456066946, 0, 0.185488589818607), Mangel.an.Gruenflaechen = c(0.0867818400964243, 
    0.071369466280513, 0.144230769230769, 0.117259414225941, 
    0.260571428571429, 0.186951433586893), Kriminalitaet = c(0.108316593009241, 
    0.083678113363674, 0.389423076923077, 0.139330543933054, 
    0.334857142857143, 0.216500877706261), Auslaender = c(0.00715146645239052, 
    0.0212039718659495, 0.0480769230769231, 0.0550209205020921, 
    0.0114285714285714, 0), Umweltbelastung = c(0.108879067898755, 
    0.0846607364501448, 0, 0.143828451882845, 0.376, 0.228203627852545
    ), Einr..f..Kinder = c(0.0693451185214946, 0.0825403392635499, 
    0.0144230769230769, 0.0527196652719665, 0, 0.0444704505558806
    ), Einr..f..Jugendliche = c(0, 0.0934526272238312, 0, 0, 
    0, 0.000877706260971328), count = c(1466, 1821, 81, 1491, 
    330, 793), cluster = c(1, 2, 3, 4, 5, 6)), .Names = c("Geschaeft", 
"Gaststaette", "Bank", "Hausarzt", "Einr..F..Aeltere", "Park", 
"Sportstaette", "OEPNV", "Mangel.an.Gruenflaechen", "Kriminalitaet", 
"Auslaender", "Umweltbelastung", "Einr..f..Kinder", "Einr..f..Jugendliche", 
"count", "cluster"), row.names = c(NA, -6L), class = "data.frame")

which I sort with

mdf <- melt(nbhpp[,-15], id.vars = 'cluster')
mdf <- transform(mdf, variable = reorder(variable, value, mean), y = cluster)

and plot with

ggplot(mdf, aes(x=variable, y=value, group=cluster, colour=factor(cluster))) + 
  geom_line() + 
  scale_y_continuous('Anteile', formatter = "percent") +
  scale_colour_hue(name='Cluster') +
  xlab('Infrastrukturmerkmal') +
  theme_bw() +
  opts(axis.text.x = theme_text(angle=90, hjust=1), legend.position = "none") +
  facet_wrap(~cluster, ncol=3)

If I understand it corre开发者_如何转开发ctly, the transform function sort the data by the average values. But how can I include these average values as a grey line to each plot?

Thanks for your help

UPDATE:

Just for clarification:

If I take a look at the output of the reorder statement

with(mdf, reorder(variable, value, mean))

than I get the following attributes:

attr(,"scores")
   Einr..f..Jugendliche        Einr..F..Aeltere              Auslaender            Sportstaette 
             0.01572172              0.01583642              0.02381364              0.02902631 
               Hausarzt                    Bank               Geschaeft         Einr..f..Kinder 
             0.03211500              0.03630061              0.04076876              0.04391644 
            Gaststaette                    Park                   OEPNV Mangel.an.Gruenflaechen 
             0.05051310              0.06799505              0.13051918              0.14452739 
        Umweltbelastung           Kriminalitaet 
             0.15692865              0.21201772 

Which are sorted in the plot from left (lowest) to right (highest). The question is, how to draw a line, with theses attributes...

Edited answer

To add a line with the cluster averages, you need to construct a data.frame that contains the data. You can extract the values from mdf:

meanscores <- attributes(mdf$variable)$scores
meandf <- data.frame(
  variable = rep(names(meanscores), 6),
  value    = rep(unname(meanscores), 6),
  cluster  = rep(1:6, each=14)
  )

Then plot using geom_line:

ggplot(mdf, aes(x=variable, y=value, group=cluster, colour=factor(cluster))) + 
  geom_line() + 
  scale_y_continuous('Anteile', formatter = "percent") +
  scale_colour_hue(name='Cluster') +
  xlab('Infrastrukturmerkmal') +
  theme_bw() +
  opts(axis.text.x = theme_text(angle=90, hjust=1), legend.position = "none") +
  facet_wrap(~cluster, ncol=3) +
  geom_line(data=meandf, aes(x=variable, y=value), colour="grey50")

Original answer

My original interpretation was that you wanted a horizontal line with overall means.

Simply add a geom_hline layer to your plot, and map the yintercept to mean(value):

ggplot(mdf, aes(x=variable, y=value, group=cluster, colour=factor(cluster))) + 
  geom_line() + 
  scale_y_continuous('Anteile', formatter = "percent") +
  scale_colour_hue(name='Cluster') +
  xlab('Infrastrukturmerkmal') +
  theme_bw() +
  opts(axis.text.x = theme_text(angle=90, hjust=1), legend.position = "none") +
  facet_wrap(~cluster, ncol=3) +
  geom_hline(aes(yintercept=mean(value)), colour="grey50")

Create a data frame containing the mean values for each group. There are many, many ways to do this in R, e.g.,

means <- ddply(mdf, .(y), summarise, mean = mean(value))

(In this case it seems that the value is the same in each facet.)

Now draw add a horizontal line to your plot. Assuming it was previously called p,

p + geom_hline(aes(yintercept = mean), data = means)