Plotting discrete predictions with probability intervals - ggplot2

Question

I need to plot some discrete predictions with probability intervals in ggplot2, but I'm having some problems.

I have the following data.frame

city    pred    min.80    max.80
BH      100     50        150
RJ      120     80        140
SP      90      80        100

I want a plot with the cities on y-axis and the predicted values on x-axis. For each discrete value of y, there should be a horizontal bar with its range being the min.80 and max.80 values. My idea is to use geom_rect from ggplot2 for doing it.

I've tried the following code, but the problem is that I'm converting the discrete variable to continuous in order to plot it, and I lose their values on the label.

> ggplot(df) + geom_rect(aes(xmin=min.80, xmax=max.80, ymin=as.numeric(city)-0.4,
+ ymax=as.numeric(city)+0.4))

Is there another way to do it?

Answer 1

I suggest you use the geom pointrange or crossbar :

ggplot(df, aes(x=city)) + 
  geom_pointrange(aes(ymin=min.80, ymax=max.80, y=pred)) +
  coord_flip()

ggplot(df, aes(x=city)) + 
  geom_crossbar(aes(ymin=min.80, ymax=max.80, y=pred)) +
  coord_flip()

Answer 2

I think you want to keep the y axis as a factor ( y=city ). This kind of (estimate+interval) data is probably is better done with something like geom_pointrange . After all, the "height" of the rectangle doesn't have an interpretation.

If you have to have the errorbars be horizontal, I've done this before in two ways:

1. using coord_flip()
2. Last time I tried coord_flip() , it was a bit limited, so I sometimes also recreated the geom_pointrange() functionality by combining geom_hline() with geom_point() .