Plotting the poisson distribution using ggplot2's stat_function

Question

I would like to plot discrete probability distributions (like the poisson distribution) using ggplot2.

I was able to plot it without using ggplot2 like this.

plot( dpois( x=0:20, lambda=1 ), type="b")

And, I was able to plot continuous probability distributions using ggplot2 like this.

ggplot(data.frame(x=c(-10,10)), aes(x)) + stat_function(fun=dnorm, args=list(0, 1))

And the code I tried was:

ggplot(data.frame(x=c(0:10)), aes(x)) + stat_function(geom="point", fun=dpois, args=list(1))

In ggplot2, How do I plot discrete probability distributions like first one?

Answer 1

The ggplot functions would have no idea where your pdf has support. If you want to plot a discrete pdf, you'll need to calculate the points yourself. And usually it makes more sense to plot these as a bar chart since it's inappropriate to interpolate probabilities between discrete values.

ggplot(transform(data.frame(x=c(0:10)), y=dpois(x, 1)), aes(x, y)) + 
    geom_bar(stat="identity")

Answer 2

stat_function will try to interpolate between the boundary values using default n=101 points. Issue with discreet distributions is that x has to hit the integer values. Try specifying n=11 in your example:

ggplot(data.frame(x=c(0:10)), aes(x)) +
    stat_function(geom="point", n=11, fun=dpois, args=list(1))

a lot simpler and much more straightforward is to use geom_point in this case:

ggplot(data.frame(x=c(0:10)), aes(x)) +
    geom_point(aes(y=dpois(x, 1)), colour="red")