We have 3 parametric bootstrapping labs on this site:
- Don't run this old lab tonight: AlgaeTopoParametricBootstrappingLab demonstrates topology test for cases in which you have one a priori group that was not recovered from your data. Does your data reject the hypothesis that the tree exists?
- SnakeTopoParametricBootstrappingLab demonstrates a very similar topology test on a larger dataset. This is built around a case in which searches on your data recover a surprising clade. Can you be confident that this surprising group is not an artifact of sampling error?
- CharacterEvoParametricBootstrappingLab demonstrates one way to use parametric bootstrapping to test a hypothesis about character evolution which depends on a phylogeny.
The AlgaeTopoParametricBootstrappingLab and SnakeTopoParametricBootstrappingLab really only differ by the type of search that can be done (the algae data set is tiny, so you can do thorough searches; the snake data set requires the use of heuristic searching) and the type of constrained search used for calculating the test statistic on the real data.
In all three of these labs, we use parsimony-based test statistics. This is rarely the correct choice for real data - we have done it for the sake of time. Parametric bootstrapping is extremely flexible - we can use the same principles to get the null distribution for any test statistic.
There is a discussion of performing other topology tests in PAUP at the end of the page the describes the tests on the Algae data set (HERE)