ggplot2 manually specifying color & linetype - duplicate legend

9
votes

I'm creating a plot with multiple linetypes, colors, and filled regions.The code below produces two legends (one shows line types, the other shows line colors) - and I need them to be combined into one legend that shows both linetype and linecolor. [there is a third legend showing the 'fill' objects, but that's fine]

I followed the methodology given here: Controlling line color and line type in ggplot legend to try to get a single legend - but ended up with this double-legend behavior - what have I done wrong?

library(ggplot2)
library(scales)
data = structure(list(Dates = structure(c(1351713600, 1351717200, 1351720800, 
  1351724400, 1351728000, 1351731600), class = c("POSIXct", "POSIXt"
  ), tzone = "MST"), CumHVAC_Def_Stoch_Min = c(146.4006, 146.6673, 
  146.9336, 147.1996, 147.4648, 147.5964), CumHVAC_Def_Stoch_1st = c(188.0087, 
  188.2753, 188.5416, 188.8077, 189.0729, 189.2045), 
  CumHVAC_Def_Stoch_Mean = c(204.7234, 204.9901, 205.2564, 205.5225, 205.7876, 205.9193), 
  CumHVAC_Def_Stoch_3rd = c(228.8813, 229.1476, 229.4135, 229.6793, 229.9442, 230.0757), 
  CumHVAC_Def_Stoch_Max = c(295.145, 295.4117, 295.6779, 295.944, 296.2092, 296.3408), 
  CumHVAC_Opt_Stoch_Min = c(112.4095, 112.6761, 112.9424, 113.2085, 113.4737, 113.6053), 
  CumHVAC_Opt_Stoch_1st = c(134.8893,135.156, 135.4223, 135.6883, 135.9535, 136.0851), 
  CumHVAC_Opt_Stoch_Mean = c(156.8854, 157.1521, 157.4184, 157.6845, 157.9496, 158.0813), 
  CumHVAC_Opt_Stoch_3rd = c(168.7301, 168.9971, 169.2636, 169.5299, 169.7953, 169.927), 
  CumHVAC_Opt_Stoch_Max = c(241.2483, 241.5151, 241.7814, 242.0476, 242.3128, 242.4444), 
  CumHVAC_Def_Dtrmn = c(188.7523,  189.0189, 189.2852, 189.5513, 189.8165, 189.9481), 

  CumHVAC_Opt_Dtrmn = c(86.8116,  87.0782, 87.3445, 87.6105, 87.8757, 88.0073),
  CS_Opt_Stoch = c(0,   0, 0, 0, 0, 0), CS_Opt_Dtrmn = c(0, 0, 0, 0, 0, 0), 
  CS_Default = c(0,   0, 0, 0, 0, 0)), .Names = c("Dates", "CumHVAC_Def_Stoch_Min", 
  "CumHVAC_Def_Stoch_1st", "CumHVAC_Def_Stoch_Mean", "CumHVAC_Def_Stoch_3rd",  
  "CumHVAC_Def_Stoch_Max", "CumHVAC_Opt_Stoch_Min", 
  "CumHVAC_Opt_Stoch_1st","CumHVAC_Opt_Stoch_Mean", "CumHVAC_Opt_Stoch_3rd", 
  "CumHVAC_Opt_Stoch_Max", "CumHVAC_Def_Dtrmn", "CumHVAC_Opt_Dtrmn", "CS_Opt_Stoch", 
  "CS_Opt_Dtrmn",    "CS_Default"), row.names = 691:696, class = "data.frame")

stochdefcolor = 'red'
stochoptcolor = 'green'
dtrmndefcolor = 'darkred'
dtrmnoptcolor = 'darkgreen'

eb09 <- aes(x = Dates, ymax = CumHVAC_Def_Stoch_3rd, ymin = CumHVAC_Def_Stoch_1st, fill="StochDef")
eb10 <- aes(x = Dates, ymax = CumHVAC_Opt_Stoch_3rd, ymin = CumHVAC_Opt_Stoch_1st, fill="StochOpt")
State = c('a','b','c','d','e','f','g','h');

ln1 <- aes(x=Dates,y=CumHVAC_Def_Stoch_Mean, color=State[1],linetype=State[1])
ln2 <- aes(x=Dates,y=CumHVAC_Opt_Stoch_Mean, color=State[2],linetype=State[2])
ln3 <- aes(x=Dates,y=CumHVAC_Def_Dtrmn,color=State[3],linetype=State[3])
ln4 <- aes(x=Dates,y=CumHVAC_Opt_Dtrmn,color=State[4],linetype=State[4])

ln5 <- aes(x=Dates,y=CumHVAC_Def_Stoch_Max,color=State[5],linetype=State[5])#,linetype = 2]
ln6 <- aes(x=Dates,y=CumHVAC_Def_Stoch_Min,color=State[6],linetype=State[6])#,linetype = 3)
ln7 <- aes(x=Dates,y=CumHVAC_Opt_Stoch_Max,color=State[7],linetype=State[7])#,linetype = 2)
ln8 <- aes(x=Dates,y=CumHVAC_Opt_Stoch_Min,color=State[8],linetype=State[8])#,linetype = 3)

quartz()
ggplot(data) + 
  geom_ribbon(eb09, alpha=0.4) +
  geom_ribbon(eb10, alpha=0.4) +
  geom_line(ln1,size=1) + 
  geom_line(ln2,size=1) + 
  geom_line(ln3,size=1) +
  geom_line(ln4,size=1) +
  geom_line(ln5,size=.7) +
  geom_line(ln6,size=.7) +
  geom_line(ln7,size=.7) +
  geom_line(ln8,size=.7) +
  xlab("X-lab") +
  ylab("Y-Lab") +
  opts(title = expression('Dummy Title'),
       panel.background = theme_rect(fill = "transparent"),
       panel.grid.minor = theme_blank(), 
       panel.grid.major = theme_blank(),
       plot.background = theme_rect(fill = "transparent")) + 
  scale_linetype_manual(values=c(1,1,1,1,2,3,2,3)) +      
  scale_colour_manual(name=c("Lines"),
                      values=c(stochdefcolor,
                               stochoptcolor,
                               dtrmndefcolor,
                               dtrmnoptcolor,
                               stochdefcolor,
                               stochdefcolor,
                               stochoptcolor,
                               stochoptcolor)) + 
  scale_fill_manual(name='1st-3rd Quartiles',
                    breaks=c('StochDef','StochOpt'),
                    values=c(stochdefcolor,stochoptcolor),
                    labels=c('Stoch DEF','Stoch OPT'))

...since I'm a new user, I can't post an image...

1
rggplot2legend
Welcome to SO and thank you for posting your code. Given that this is not a straightforward question, it would be helpful if you could include your data (the data object above) or a a subset thereof, perhaps by using dput as the beginning. If you do that other users will be able to copy, paste and experiment with your code on their own R setups. – SlowLearner
That's a lot to wade through, but if I had to guess, your problem is that you are forcing ggplot to create a bunch of manual scales, rather than simply adding State as a variable and mapping color and linetype to it. That would involve some melting and rearranging of your data, but I'm 99% certain you can do this with only one geom_line call. – joran
@SlowLearner - Thanks for the tip, I'm adding data and enough code to the posting so that it should run for anyone as it does for me. – RyanStochastic
@joran Given what I've seen in related postings, I don't want ggplot to choose the colors for me. I want to specify the colors myself to make the graph more readable. Most of the data is related in one way or another, and I'm using similar colors and/or linetypes to indicate that different lines correspond to similar subsets of data. (i.e., any data that has 'Def' in the variable name should be red, any data that has 'Opt' in the variable name should be green. Any variable that has 'stoch' in its name should be light, any variable that has 'dtrmn' should be dark...and so on) – RyanStochastic
If you play with the code provided below, you will see that nothing I suggested precludes you from selecting your own values for the scales. – joran

1 Answers

13
votes

As said in the comment by @joran You need to create a variable state and bind it to color and linetype then ggplot do the rest for you. For example ,Here the plot s very simple since you put data in the right shape.

To manipulate data you need I advise you to learn plyr and reshape2 !

## I use your data 
## I melt my data 
library(reshape2)
measure.vars  <- colnames(dat)[c(1:12)][-c(1,3,5,10)]
mydata.melted <- melt(mydata, 
                      measure.vars= measure.vars, ## I don't use all the variables only
                                                                      States ones
                      id.vars='Dates')


states.list <- list('a','b','c','d','e','f','g','h')
names(states.list) <- measure.vars
mydata.melted$State <- NA
library(plyr)
mydata.melted <- ddply(mydata.melted,
                       .(variable),transform,
                       State=states.list[[unique(variable)]])
## I plot using the rights aes    

stochdefcolor = 'red'
stochoptcolor = 'green'
dtrmndefcolor = 'darkred'
dtrmnoptcolor = 'darkgreen'
library(ggplot2)
ggplot(subset(mydata.melted)) + 
  geom_line(aes(x=Dates,y=value,
                color=State,linetype=State))+
  scale_linetype_manual(values=c(1,1,1,1,2,3,2,3)) +  
  scale_size_manual(values =rep(c(1,0.7),each=4))+
  scale_color_manual(values=c(stochdefcolor,stochoptcolor,
                               dtrmndefcolor, dtrmnoptcolor,
                               stochdefcolor,stochdefcolor,
                               stochoptcolor,stochoptcolor))

enter image description here