Adding confidence intervals ggplot

Question

I want to plot intervals of confidence of fitted values. I read some post related, but I am still stuck..These are a sample of the date:

pd <-structure(list(date = 1:5, obs = c(44.6651011845397, 62.3441339250369, 
          52.8968240161506, 51.7795930633478, 63.1284636561025), pred = c(47.2643891039645, 
          55.7996450577325, 52.9566469533233, 51.3393289316, 59.0011440099732)), 
          .Names = c("date", "obs", "pred"), row.names = c(NA, 5L), class = "data.frame")

pd2 <- structure(list(date = 1:5, lwr = c(44.8529592578518, 54.9926370476338, 
         51.7358955911624, 49.401869166722, 58.1674619447108), upr = c(49.6758189500772, 
         56.6066530678312, 54.1773983154842, 53.2767886964779, 59.8348260752356
          )), .Names = c("date", "lwr", "upr"), row.names = c(NA, 5L), class = "data.frame")

dd  <-  melt(pd, id=c("date"))   #Data
dd2 <- melt(pd2,id=c("date"))    #Intervals of conf.
p <- ggplot(dd) + geom_line(aes(x=date, y=value, colour=variable)) 
p <- p + geom_smooth(aes(x=date, y=value, ymax=lwr, ymin=upr), #1 way
            colour='grey', data=dd2, stat='identity')

Also tried...

# p+ geom_ribbon(data=dd2,aes(ymin=lwr,ymax=upr),alpha=0.3) #2.

I received the error: Error in eval(expr, envir, enclos) : object 'lwr' not found ....what am I missing?

I also tried to do it without using melt ... but then I also had problems with the legend!

Nikos Nikos · Accepted Answer · 2014-11-19T12:52:28

In the first way, the dd object you are using is not having the lwr/upr columns. So you cannot really plot them.

Can you do a:

dd<-merge(dd,pd2,by='date')

just after the melts and then:

p <- ggplot(dd) + geom_line(aes(x=date, y=value, colour=variable)) 
p + geom_ribbon(data=dd,aes(x=date, y=value, ymin=lwr,ymax=upr, group=variable),alpha=0.3)

Is this helping?

Adding confidence intervals ggplot

1 Answers