# BEAST Tutorial

There are two versions of BEAST: BEAST1, and BEAST2. Their interfaces are very similar, and I'm slowly migrating toward BEAST2, because it can do everything that BEAST1 can do, and more.

### The Best Quickstart Tutorial I Know Of

This is somewhat a cheat - I'm not actually writing the tutorial, but dropping a link to an already excellent tutorial written by Trevor Bedford. That tutorial gives a demo of a whole analysis pipeline from `BEAUTI`

through to `Tracer`

, `Figtree`

and `TreeAnnotator`

. Ultimately, though, the more difficult question is about the most appropriate choice of parameters to select for the given study.

This set of Taming-the-Beast(2) tutorials are also useful, though applicable only to BEAST 2.

### Useful Links

- The BEAST2 tutorials
- The exercises and tutorials from the Taming the Beast 2017 workshop are also good.
- A temporary tutorial on tip-dating in BEAST2, written by Louis du Plessis.

### Pet Peeve

If I read the documentation correctly, BEAST provides some statistical support for the quality of the tree by recomputing the tree over and over again from the same input - they tend to call this "bootstrap support". This is *not* called bootstrap, it's called Monte Carlo. Bootstrap in statistics strictly refers to *sampling with replacement*. A Monte Carlo simulation is a procedure where, say, you wish to find out the probability of heads on a weighted coin (so not necessarily 0.5). Flip the coin 1000 times; the proportion of the number of times that the coin comes up heads will be the empirical probability of heads. That's a Monte Carlo simulation.