Why does the ggplot legend show the “colour” parameter?

Question

I have the following example:

data <- structure(list(a = c(-1.25549186262767, -0.840855480786298, -
0.635371312524283, 
-0.602907981454667, -0.472166385166945, -0.285773634866154, 0.0701427664273268, 
0.138108224803923, 1.38435934347858, 1.71144087270237), b = c(-3.44400412039417, 
0.675644682353751, -1.04793816522475, -7.38303347186651, 2.34519166466874, 
0.334780748573386, 4.76806919070976, 4.8633533150074, 3.50106026731172, 
-1.27172351054143), c = c(-3.02376206439776, -2.56390769080574, 
-1.48659913867609, -1.27976202274701, -0.368725655874139, 1.08537150160227, 
3.98619381956471, 4.50687017428731, 4.10341582203292, -1.61769414438858
), d = c(5.71851494232005, 2.90539833491649, 2.75195159216204, 
2.73478241733301, 2.65941820902101, 2.60630235726839, 3.34836154776286, 
3.62938300664006, 4.61153521538016, 5.56230567213863), e = c(8.98703236551896, 
4.5660296657415, 4.32487774825464, 4.29789523068949, 4.17945528847841, 
4.09598014088541, 5.26217626511884, 5.70382046327322, 7.24733897758039, 
8.74153894964533)), .Names = c("a", "b", "c", "d", "e"), row.names = c(NA, 
-10L), class = "data.frame")

ggplot(data, aes(x=a, y=b)) + geom_point() + 
    geom_line(aes(x=a, y=c)) + 
    geom_line(aes(x=a, y=(c - d), colour="red")) +
    geom_line(aes(x=a, y=(c + d), colour="red")) +
    geom_line(aes(x=a, y=(c - e), colour="blue")) +
    geom_line(aes(x=a, y=(c + e), colour="blue")) 

I want the labels to be "d" and "e", but they are instead being given the value of the "colour" field. Two questions:

1) How can I eliminate the legend entirely? 2) If I want to include the legend, how can I have it set to specific values rather than the colour?

Answer 1

The important thing to remember here is that items inside of aes() will map data to aesthetics. If you are setting a constant value(s), then you can move that information outside of aes(). A slight modification of your code yields:

ggplot(data, aes(x=a, y=b)) + geom_point() + 
geom_line(aes(x=a, y=c)) + 
geom_line(aes(x=a, y=(c - d)), colour="red") +
geom_line(aes(x=a, y=(c + d)), colour="red") +
geom_line(aes(x=a, y=(c - e)), colour="blue") +
geom_line(aes(x=a, y=(c + e)), colour="blue") 

Which gives you what you are after without any legend. I like @koshke's approach for including a legend above so won't duplicate that. The other approach you could take would be to perform your data manipulation outside of the call to ggplot2() and then melt() it into long format before plotting. That would shorten your call to ggplot() since you could get rid of the multiple calls to geom_line(), but there's obviously the overhead of preprocessing the data. Probably 6 in one, 1/2 dozen in the other for this problem, but something to keep in mind for future problems.

Answer 2

is this what you want to do?

ggplot(data, aes(x=a, y=b)) + geom_point() + 
    geom_line(aes(x=a, y=c)) + 
    geom_line(aes(x=a, y=(c - d), colour="d")) +
    geom_line(aes(x=a, y=(c + d), colour="d")) +
    geom_line(aes(x=a, y=(c - e), colour="e")) +
    geom_line(aes(x=a, y=(c + e), colour="e")) +
    scale_colour_manual(name="legend title", 
                        values=c("red", "blue"),
                        breaks=c("d", "e"))

Answer 3

Eliminating the the label could be done via adding + opts(legend.position = "none") option also.