Like the t-test and family of t-distributions, the F-test has a family of F-distributions
The family of F-distributions depends on:
- the Number of subjects per group
- the Number of groups
ANOVA can handle as many groups as you like. So there are different sampling distributions. This different sampling distributions can be generated in R.
Each different curve varies depending on the number of groups and number of people in a group.
It's not symmetrical because negative values are impossible. It's a ratio of variances. The expected value under the null hypothesis is just one.
It's the same concept as t and z distributions but it just looks a little different.