CODEWITHSUNDEEP
rggplot2
mgilbert
mgilbert

Reputation: 3665

Mixing line and scatterplot in ggplot

I've looked around a fair bit but I am stumped on this one. I cannot find a way to plot a line that is unassociated with a scatterplot. Here is some of my data and code to clarify the problem. I have data of the following form

> head(allData)
  AnnounceDate MarketProbability DealStatus binary BrierScore
1   2000-04-10         0.3333333   Complete      1  0.2340565
2   2000-06-14         0.2142857   Complete      1  0.3618200
3   2000-06-26         0.6846154   Complete      1  0.3690167
4   2000-06-16         0.1875000   Complete      1  0.4364041
5   2000-10-05         0.9555556   Complete      1  0.3078432
6   2000-10-19         0.8500000   Complete      1  0.2670799

I would like to create a scatter plot of the MarketProbabilities vs. the AnnounceDate, and identify whether a DealStatus is Completed or Terminated using colour.

(ggplot(data=allData, aes(x=AnnounceDate, y=MarketProbability, colour=DealStatus))
 + geom_point() + scale_colour_hue(h = c(180,0)))

enter image description here

I would also like to overlay a rolling Brier score, I have tried

(ggplot(data=allData, aes(x=AnnounceDate, y=MarketProbability, colour=DealStatus))
 + geom_point() + scale_colour_hue(h=c(180,0))
 + geom_line(aes(x=AnnounceDate, y=BrierScore)))

enter image description here

I am confused why two coloured lines are plotted? How do I dissociate the line with the previous plotted data?

Upvotes: 9

Views: 16614

Answers (2)

MattBagg
MattBagg

Reputation: 10478

The aes() in ggplot() are inherited by the subsequent geoms. You assigned colour=DealStatus, which gets inherited by geom_line() and it needs to make two lines to make two colors. I can't test this because I am on my phone, but try moving the colour assignment out of the main ggplot and into the geom_point():

ggplot(data = allData, aes(x=AnnounceDate)) +
     geom_point(aes(colour=DealStatus, y=MarketProbability)) + 
     scale_colour_hue(h = c(180,0)) +
     geom_line(aes(y=BrierScore, group=1))

I am adding a constant as a group argument to tell it to connect all the points. Otherwise, ggplot sometimes makes the wrong guesses.

Upvotes: 5

bdemarest
bdemarest

Reputation: 14665

As @MattBagg has pointed out, this issue is fixed by moving the aes() for colour out of the main ggplot() call and into the geom_point() call.

library(ggplot2)

allData = read.table(header=TRUE,
colClasses=c("Date", "numeric", "character", "numeric", "numeric"),
text="AnnounceDate MarketProbability DealStatus binary BrierScore
2000-04-10         0.3333333   Complete      1  0.2340565
2000-06-14         0.2142857   Complete      1  0.3618200
2000-06-26         0.6846154   Complete      1  0.3690167
2000-06-16         0.1875000   Complete      1  0.4364041
2000-10-05         0.9555556   Complete      1  0.3078432
2000-10-19         0.8500000   Complete      1  0.2670799")

p1 = ggplot(data=allData, aes(x=AnnounceDate)) +
     geom_point(aes(y=MarketProbability, colour=DealStatus)) + 
     scale_colour_hue(h = c(180,0)) + 
     geom_line(aes(y=BrierScore))

ggsave(filename="plot_1.png", plot=p1, height=2.5, width=5)

enter image description here

Upvotes: 7

Related Questions