Plot the confidence band with ggplot2

Question

I have a dataset that is similar to this:

x <- data.frame(date = c(20190902, 20190903, 20190904),
            Group = c(rep("A", 3)),
            mean = c(2.5, 3.4, 4.6),
            ci_upper = c(1.2, 0.5, 0.3),
            ci_lower = c(0.5, 0.4, 0.25))

y <- data.frame(date= c(20190902, 20190903, 20190904),
            Group = c(rep("B", 3)),
            mean = c(0.4, 3.8, 6.2),
            ci_upper = c(1.9, 0.9, 0.5),
            ci_lower = c(0.5, 0.8, 0.8))

df <- rbind(x, y)

I would like to plot the confidence band across the timeframe, with 2 different groups (A and B).

Currently I'm using this method but didn't work:

p <- ggplot(df) + 
    geom_line(aes(y = mean, x = date, group = type ))+
    geom_ribbon(aes(ymin = ci_lower, ymax = ci_upper, x = week, fill = "grey70"), alpha = 0.3)+
    scale_colour_manual("", values = "blue")+
    scale_fill_manual("", values = "grey12")

I wasn't sure how I can approach this.

Answer 1

You are almost there. Only some small corrections of the aes() are needed.

But first I would slightly modify the input just to make the result looking prettier (now the ci_upper / ci_lower are not always more/less as compared with a corresponding mean value):

# to ensure reproducibility of the samples
set.seed(123)
df$ci_lower <- df$mean - sample(nrow(x))
df$ci_upper <- df$mean + sample(nrow(x))

The main thing which should be changed in your ggplot() call is definition of the aesthetics which will be used for plotting. Note, please, that default aesthetics values should be set only once.

p <- ggplot(df, 
    aes(x = as.Date(as.character(date), format = "%Y%m%d"), 
        y = mean,
        group = Group, col = Group, fill = Group)) + 
    geom_line() +
    geom_ribbon(aes(ymin = ci_lower, ymax = ci_upper), alpha = 0.3)+
    scale_colour_manual("", values = c("red", "blue")) +
    scale_fill_manual("", values = c("red", "blue"))

The result is as follows:

Actually, the last two code rows are even not necessary, as the default ggplot-color scheme (which you have used to show the desired result) looks very nice, also.